public void setDefaultVoice(Voice voice) {
if (voice == null) {
return;
}
// check that voice locale fits before accepting the voice:
Locale voiceLocale = null;
voiceLocale = voice.getLocale();
Locale docLocale = getLocale();
if (docLocale == null
&& getType().isXMLType()
&& getDocument() != null
&& getDocument().getDocumentElement().hasAttribute("xml:lang")) {
docLocale =
MaryUtils.string2locale(getDocument().getDocumentElement().getAttribute("xml:lang"));
}
if (docLocale != null
&& voiceLocale != null
&& !(MaryUtils.subsumes(docLocale, voiceLocale)
|| MaryUtils.subsumes(voiceLocale, docLocale))) {
logger.warn(
"Voice `"
+ voice.getName()
+ "' does not match document locale `"
+ docLocale
+ "' -- ignoring!");
}
this.defaultVoice = voice;
}
public Mary4To5VoiceConverter(List<VoiceComponentDescription> voiceDescriptions, File voiceZip) {
voiceDescription = null;
mary4Zip = voiceZip;
for (VoiceComponentDescription d : voiceDescriptions) {
if (d.getPackageFilename().equals(mary4Zip.getName())) {
voiceDescription = d;
break;
}
}
if (voiceDescription == null) {
throw new IllegalArgumentException(
"No matching voice description for file " + mary4Zip.getName());
}
if (!MaryUtils.isLog4jConfigured()) {
BasicConfigurator.configure();
}
logger = Logger.getLogger(this.getClass());
logger.info(
voiceDescription.getName()
+ " "
+ voiceDescription.getVersion()
+ " ("
+ voiceDescription.getLocale()
+ " "
+ voiceDescription.getGender()
+ ")");
}
/**
* Phonemise the word text. This starts with a simple lexicon lookup, followed by some heuristics,
* and finally applies letter-to-sound rules if nothing else was successful.
*
* @param text the textual (graphemic) form of a word.
* @param pos the part-of-speech of the word
* @param g2pMethod This is an awkward way to return a second String parameter via a
* StringBuilder. If a phonemisation of the text is found, this parameter will be filled with
* the method of phonemisation ("lexicon", ... "rules").
* @return a phonemisation of the text if one can be generated, or null if no phonemisation method
* was successful.
*/
public String phonemise(String text, String pos, StringBuilder g2pMethod) {
// First, try a simple userdict and lexicon lookup:
String result = userdictLookup(text, pos);
if (result != null) {
g2pMethod.append("userdict");
return result;
}
result = lexiconLookup(text, pos);
if (result != null) {
g2pMethod.append("lexicon");
return result;
}
// HB 150915 adding secondary lexicon
result = secondary_lexiconLookup(text, pos);
if (result != null) {
g2pMethod.append("lexicon");
return result;
}
// Lookup attempts failed. Try normalising exotic letters
// (diacritics on vowels, etc.), look up again:
String normalised = MaryUtils.normaliseUnicodeLetters(text, getLocale());
if (!normalised.equals(text)) {
result = userdictLookup(normalised, pos);
if (result != null) {
g2pMethod.append("userdict");
return result;
}
result = lexiconLookup(normalised, pos);
if (result != null) {
g2pMethod.append("lexicon");
return result;
}
// HB 150915 adding secondary lexicon
result = secondary_lexiconLookup(text, pos);
if (result != null) {
g2pMethod.append("lexicon");
return result;
}
}
// Cannot find it in the lexicon -- apply letter-to-sound rules
// to the normalised form
String phones = lts.predictPronunciation(text);
try {
result = lts.syllabify(phones);
} catch (IllegalArgumentException e) {
logger.error(String.format("Problem with token <%s> [%s]: %s", text, phones, e.getMessage()));
}
if (result != null) {
g2pMethod.append("rules");
return result;
}
return null;
}
/**
* For an element in a MaryXML document, do what you can to determine the appropriate
* AllophoneSet. First search for the suitable voice, then if that fails, go by locale.
*
* @param e
* @return an allophone set if there is any way of determining it, or null.
* @throws MaryConfigurationException if a suitable allophone set exists in principle, but there
* were problems loading it.
*/
public static AllophoneSet determineAllophoneSet(Element e) throws MaryConfigurationException {
AllophoneSet allophoneSet = null;
Element voice = (Element) MaryDomUtils.getAncestor(e, MaryXML.VOICE);
Voice maryVoice = Voice.getVoice(voice);
if (maryVoice == null) {
// Determine Locale in order to use default voice
Locale locale =
MaryUtils.string2locale(
e.getOwnerDocument().getDocumentElement().getAttribute("xml:lang"));
maryVoice = Voice.getDefaultVoice(locale);
}
if (maryVoice != null) {
allophoneSet = maryVoice.getAllophoneSet();
} else {
Locale locale =
MaryUtils.string2locale(
e.getOwnerDocument().getDocumentElement().getAttribute("xml:lang"));
allophoneSet = determineAllophoneSet(locale);
}
return allophoneSet;
}
/**
* Instantiate an object by calling one of its constructors.
*
* @param objectInitInfo a string description of the object to instantiate. The objectInitInfo is
* expected to have one of the following forms:
* <ol>
* <li>my.cool.Stuff
* <li>my.cool.Stuff(any,string,args,without,spaces)
* <li>my.cool.Stuff(arguments,$my.special.property,other,args)
* </ol>
* where 'my.special.property' is a property in one of the MARY config files.
* @return the newly instantiated object.
* @throws ClassNotFoundException
* @throws IllegalArgumentException
* @throws InstantiationException
* @throws IllegalAccessException
* @throws InvocationTargetException
* @throws SecurityException
* @throws NoSuchMethodException
*/
public static Object instantiateObject(String objectInitInfo) throws MaryConfigurationException {
Object obj = null;
String[] args = null;
String className = null;
try {
if (objectInitInfo.contains("(")) { // arguments
int firstOpenBracket = objectInitInfo.indexOf('(');
className = objectInitInfo.substring(0, firstOpenBracket);
int lastCloseBracket = objectInitInfo.lastIndexOf(')');
args = objectInitInfo.substring(firstOpenBracket + 1, lastCloseBracket).split(",");
for (int i = 0; i < args.length; i++) {
if (args[i].startsWith("$")) {
// replace value with content of property named after the $
args[i] = MaryProperties.getProperty(args[i].substring(1));
}
args[i] = args[i].trim();
}
} else { // no arguments
className = objectInitInfo;
}
Class<? extends Object> theClass = Class.forName(className).asSubclass(Object.class);
// Now invoke Constructor with args.length String arguments
if (args != null) {
Class<String>[] constructorArgTypes = new Class[args.length];
Object[] constructorArgs = new Object[args.length];
for (int i = 0; i < args.length; i++) {
constructorArgTypes[i] = String.class;
constructorArgs[i] = args[i];
}
Constructor<? extends Object> constructor =
(Constructor<? extends Object>) theClass.getConstructor(constructorArgTypes);
obj = constructor.newInstance(constructorArgs);
} else {
obj = theClass.newInstance();
}
} catch (Exception e) {
// try to make e's message more informative if possible
throw new MaryConfigurationException(
"Cannot instantiate object from '"
+ objectInitInfo
+ "': "
+ MaryUtils.getFirstMeaningfulMessage(e),
e);
}
return obj;
}
/**
* Apply the SoP to a Target to get its predicted value
*
* @param target target from where to predict
* @return result predicted value
*/
@Override
protected float evaluate(Target target) {
float result = 0;
if (targetAttributeName.contentEquals("f0")) {
result = (float) sopModels.get("f0").interpret(target);
} else {
if (target.getAllophone().isVowel())
result = (float) sopModels.get("vowel").interpret(target);
else if (target.getAllophone().isConsonant())
result = (float) sopModels.get("consonant").interpret(target);
else if (target.getAllophone().isPause())
result = (float) sopModels.get("pause").interpret(target);
else {
// ignore but complain
MaryUtils.getLogger("SoPModel")
.warn("Warning: No SoP model for target " + target.toString());
}
}
return result;
}
/**
* Set up a new wagon process. Wagon.setWagonExecutable() must be called beforehand.
*
* @param featureDefinition
* @param featureVectors
* @param aDistanceMeasure
* @param dir
* @param balance
* @param stop
* @throws IOException if there was no call to Wagon.setWagonExecutable() with an executable file
* before calling this constructor.
*/
public Wagon(
String id,
FeatureDefinition featureDefinition,
FeatureVector[] featureVectors,
DistanceMeasure aDistanceMeasure,
File dir,
int balance,
int stop)
throws IOException {
if (wagonExecutable == null || !wagonExecutable.isFile()) {
throw new IOException("No wagon executable set using Wagon.setExecutable()!");
}
this.logger = MaryUtils.getLogger("Wagon");
this.id = id;
this.featureDefinition = featureDefinition;
this.fv = featureVectors;
this.distMeasure = aDistanceMeasure;
this.distFile = new File(dir, id + ".dist");
this.descFile = new File(dir, id + ".desc");
this.featFile = new File(dir, id + ".feat");
this.cartFile = new File(dir, id + ".cart");
this.systemCall =
wagonExecutable.getAbsolutePath()
+ " -desc "
+ descFile.getAbsolutePath()
+ " -data "
+ featFile.getAbsolutePath()
+ " -balance "
+ balance
+ " -distmatrix "
+ distFile.getAbsolutePath()
+ " -stop "
+ stop
+ " -output "
+ cartFile.getAbsolutePath();
}
/**
* Verify if the java virtual machine is in a very low memory condition. The memory is considered
* very low if less than half a specified value is still available for processing. "Available"
* memory is calculated using <code>availableMemory()</code>.The threshold value can be specified
* as the Mary property mary.lowmemory (in bytes). It defaults to 20000000 bytes.
*
* @return a boolean indicating whether or not the system is in very low memory condition.
*/
public static boolean veryLowMemoryCondition() {
return MaryUtils.availableMemory() < lowMemoryThreshold() / 2;
}
public MaryData process(MaryData d) throws Exception {
Document doc = MaryXML.newDocument();
Element root = doc.getDocumentElement();
Locale locale = d.getLocale();
if (locale != null) {
root.setAttribute("xml:lang", MaryUtils.locale2xmllang(locale));
}
Element paragraph = MaryXML.appendChildElement(root, MaryXML.PARAGRAPH);
List<Utterance> utterances = d.getUtterances();
Iterator<Utterance> it = utterances.iterator();
while (it.hasNext()) {
Utterance utterance = it.next();
Element insertHere = paragraph;
if (logger.getEffectiveLevel().equals(Level.DEBUG)) {
StringWriter sw = new StringWriter();
PrintWriter pw = new PrintWriter(sw);
utterance.dump(pw, 2, name(), true); // padding, justRelations
logger.debug("Converting the following Utterance to XML:\n" + sw.toString());
}
// Make sure we have the correct voice:
Voice maryVoice = null;
if (utterance.getVoice() != null) {
maryVoice = FreeTTSVoices.getMaryVoice(utterance.getVoice());
}
if (maryVoice != null) {
if (insertHere.getTagName().equals(MaryXML.VOICE)) {
// Are utterance voice and voiceElement voice the same?
if (maryVoice.hasName(insertHere.getAttribute("name"))) {
// then insertHere is set OK, leave it like it is
} else {
// get one higher up, create new voice element after this
// one, and make insertHere point to the new voice element
Element parent = (Element) insertHere.getParentNode();
Element newVoice = MaryXML.createElement(doc, MaryXML.VOICE);
parent.appendChild(newVoice);
newVoice.setAttribute("name", maryVoice.getName());
insertHere = newVoice;
}
} else {
// Check if the last child of insertHere is a voice with the right name
Element lastChild = MaryDomUtils.getLastChildElement(insertHere);
if (lastChild != null
&& lastChild.getTagName().equals(MaryXML.VOICE)
&& maryVoice.hasName(lastChild.getAttribute("name"))) {
insertHere = lastChild;
} else {
// create a new voice element, insert it as a child of this
// node, and let insertHere point to it
Element newVoice = MaryXML.createElement(doc, MaryXML.VOICE);
insertHere.appendChild(newVoice);
newVoice.setAttribute("name", maryVoice.getName());
insertHere = newVoice;
}
}
// Now insertHere is the correct <voice> element.
// Any prosodic settings to insert?
Element prosody = insertProsodySettings(insertHere, utterance);
if (prosody != null) insertHere = prosody;
}
// Create a sentence element <s> for this utterance:
Element sentence = MaryXML.createElement(doc, MaryXML.SENTENCE);
insertHere.appendChild(sentence);
fillSentence(sentence, utterance);
}
if (logger.getEffectiveLevel().equals(Level.DEBUG)) {
logger.debug("Constructed the following XML structure:");
MaryNormalisedWriter mnw = new MaryNormalisedWriter();
ByteArrayOutputStream debugOut = new ByteArrayOutputStream();
mnw.output(doc, debugOut);
logger.debug(debugOut.toString());
}
MaryData output = new MaryData(outputType(), d.getLocale());
output.setDocument(doc);
return output;
}
/**
* Model for predicting duration and F0 from HMMs
*
* @author marcela
*/
public class HMMModel extends Model {
/** Configuration information of the model */
private HMMData htsData = null;
/** HMM trees and pdfs for this model. */
private CartTreeSet cart;
/** Feature definition used when training HMMs. */
FeatureDefinition hmmFeatureDefinition;
/** to calculate duration in seconds. */
private float fperiodsec;
protected static Logger logger = MaryUtils.getLogger("HMMModel");
/**
* If the model is instantiated because the same HHMModel is used for predicting both F0 and
* duration, set this variable true; this is done in Voice.loadAcousticModels(), when creating the
* models.
*/
private boolean predictDurAndF0 = false;
/**
* This list keeps a copy of the utterance model, this is done when the same HMMModel is used for
* predicting durations and F0, the idea is to keep in the utterance model list the state
* durations predicted together with duration, these state durations are used when predicting F0,
* so the same state duration is applied.
*/
private Map<List<Element>, HTSUttModel> uttModels = new WeakHashMap<List<Element>, HTSUttModel>();
/**
* Model constructor
*
* @param featureManager the feature processor manager used to compute the symbolic features used
* for prediction
* @param dataFileName in HMM models this data file corresponds to the configuration file of the
* HMM voice
* @param targetAttributeName attribute in MARYXML to predict
* @param targetAttributeFormat print style, not used in HMM models
* @param featureName not used in HMMModel
* @param predictFrom not used in HMMModel
* @param applyTo not used in HMMModel
* @throws MaryConfigurationException if there are missing files or problems loading trees and pdf
* files.
*/
public HMMModel(
FeatureProcessorManager featureManager,
String voiceName,
InputStream dataStream,
String targetAttributeName,
String targetAttributeFormat,
String featureName,
String predictFrom,
String applyTo)
throws MaryConfigurationException {
super(
featureManager,
voiceName,
dataStream,
targetAttributeName,
targetAttributeFormat,
featureName,
predictFrom,
applyTo);
if (!(targetAttributeName.contentEquals("d") || targetAttributeName.contentEquals("f0"))) {
throw new MaryConfigurationException(
"targetAttributeName = " + targetAttributeName + " Not known");
}
load();
}
/**
* This variable is set to true whenever the same HMMModel is used to predict both duration and
* F0. by default the variable is false, so that means that two different HMMModels are used for
* predicting duration and F0, in this case there is no state durations information to predict F0.
*
* @param bval
*/
public void setPredictDurAndF0(boolean bval) {
predictDurAndF0 = bval;
}
/**
* Load trees and pdfs, from HMM configuration file.
*
* @throws MaryConfigurationException if there are missing files or problems loading trees and pdf
* files.
*/
@Override
protected void loadData() throws IOException, MaryConfigurationException {
if (htsData == null) htsData = new HMMData();
// we use the configuration of the HMM voice whose hmm models will be used
htsData.initHMMDataForHMMModel(voiceName);
cart = htsData.getCartTreeSet();
fperiodsec = ((float) htsData.getFperiod() / (float) htsData.getRate());
predictionFeatureNames = htsData.getFeatureDefinition().getFeatureNames();
}
/**
* Predict duration for the list of elements. If the same HMMModel is used to predict duration and
* F0 then a utterance model is created and kept in a WeakHashMap, so the next call to this
* module, for predicting F0, can use that utterance model.
*
* @param elements elements from MaryXML for which to predict the values
* @throws MaryConfigurationException if error searching in HMM trees.
*/
@Override
public void applyTo(List<Element> elements) throws MaryConfigurationException {
logger.debug("predicting duration");
HTSUttModel um = predictAndSetDuration(elements, elements);
if (predictDurAndF0) { // this same model will be used for predicting F0 -- remember um
uttModels.put(elements, um);
}
}
/**
* Predict F0 for the list of elements and apply to another list of elements. If the same HMMModel
* is used to predict duration and F0 then there must be a utterance model created in a previous
* call to this module, that will be used to predict F0. If there is no previously created
* utterance model then one is created.
*
* @param predictFromElements elements from MaryXML for which to predict the values
* @param applyToElements elements from MaryXML for which to apply the predicted values
* @throws MaryConfigurationException if error searching in HMM trees.
*/
@Override
public void applyFromTo(List<Element> predictFromElements, List<Element> applyToElements)
throws MaryConfigurationException {
logger.debug("predicting F0");
// Two possibilities: Either we have an uttModel due to a previous call to applyTo()
// (in which case the lookup key should be applyToElements),
// or we don't -- in which case we must create an uttModel from the XML.
HTSUttModel um;
if (predictDurAndF0) {
logger.debug("using already created utterance model, it contains predicted state durations.");
um =
uttModels.get(
applyToElements); // it must be already created so get it from the uttModels Map
} else {
logger.debug("creating utterance model with equal values for state durations.");
um =
createUttModel(
predictFromElements); // create a um, state durations are set equal for all states
}
assert um != null;
predictAndSetF0(applyToElements, um);
}
/**
* Predict durations and state durations from predictFromElements and apply durations to
* applyToElements. A utterance model is created that contains the predicted state durations.
*
* @param predictFromElements elements to predict from
* @param applyToElements elements to apply predicted durations
* @return HTSUttModel a utterance model
* @throws MaryConfigurationException if error searching in HMM trees.
*/
private HTSUttModel predictAndSetDuration(
List<Element> predictFromElements, List<Element> applyToElements)
throws MaryConfigurationException {
List<Element> predictorElements = predictFromElements;
List<Target> predictorTargets = getTargets(predictorElements);
FeatureVector fv = null;
HTSUttModel um = new HTSUttModel();
FeatureDefinition feaDef = htsData.getFeatureDefinition();
double diffdurOld = 0.0;
double diffdurNew = 0.0;
String durAttributeName = "d";
try {
// (1) Predict the values
for (int i = 0; i < predictorTargets.size(); i++) {
fv = predictorTargets.get(i).getFeatureVector();
um.addUttModel(new HTSModel(cart.getNumStates()));
HTSModel m = um.getUttModel(i);
/* this function also sets the phone name, the phone between - and + */
m.setPhoneName(fv.getFeatureAsString(feaDef.getFeatureIndex("phone"), feaDef));
/* Check if context-dependent gv (gv without sil) */
if (htsData.getUseContextDependentGV()) {
if (m.getPhoneName().contentEquals("_")) m.setGvSwitch(false);
}
/* increment number of models in utterance model */
um.setNumModel(um.getNumModel() + 1);
/* update number of states */
um.setNumState(um.getNumState() + cart.getNumStates());
// Estimate state duration from state duration model (Gaussian)
diffdurNew = cart.searchDurInCartTree(m, fv, htsData, diffdurOld);
diffdurOld = diffdurNew;
double duration = m.getTotalDur() * fperiodsec; // in seconds
um.setTotalFrame(um.getTotalFrame() + m.getTotalDur());
// System.out.format("HMMModel: phone=%s duration=%.3f sec. m.getTotalDur()=%d\n",
// m.getPhoneName(), duration, m.getTotalDur());
/* Find pdf for LF0, this function sets the pdf for each state.
* and determines, according to the HMM models, whether the states are voiced or unvoiced, (it can be possible that some states are voiced
* and some unvoiced).*/
cart.searchLf0InCartTree(m, fv, feaDef, htsData.getUV());
for (int mstate = 0; mstate < cart.getNumStates(); mstate++) {
for (int frame = 0; frame < m.getDur(mstate); frame++)
if (m.getVoiced(mstate)) um.setLf0Frame(um.getLf0Frame() + 1);
}
// set the value in elements
Element element = applyToElements.get(i);
// "evaluate" pseudo XPath syntax:
// TODO this needs to be extended to take into account targetAttributeNames like "foo/@bar",
// which would add the
// bar attribute to the foo child of this element, creating the child if not already
// present...
if (durAttributeName.startsWith("@")) {
durAttributeName = durAttributeName.replaceFirst("@", "");
}
String formattedTargetValue = String.format(targetAttributeFormat, duration);
// System.out.println("HMMModel: formattedTargetValue = " + formattedTargetValue);
// if the attribute already exists for this element, append targetValue:
if (element.hasAttribute(durAttributeName)) {
formattedTargetValue =
element.getAttribute(durAttributeName) + " " + formattedTargetValue;
}
// set the new attribute value:
element.setAttribute(durAttributeName, formattedTargetValue);
}
return um;
} catch (Exception e) {
throw new MaryConfigurationException("Error searching in tree when predicting duration. ", e);
}
}
/**
* Predict F0 from the utterance model and apply to elements
*
* @param applyToElements elements to apply predicted F0s
* @param um utterance model that contains the set of elements (phonemes) and state durations for
* generating F0.
* @throws MaryConfigurationException if error generating F0 out of HMMs trees and pdfs.
*/
private void predictAndSetF0(List<Element> applyToElements, HTSUttModel um)
throws MaryConfigurationException {
HTSModel m;
try {
String f0AttributeName = "f0";
HTSParameterGeneration pdf2par = new HTSParameterGeneration();
/* Once we have all the phone models Process UttModel */
/* Generate sequence of speech parameter vectors, generate parameters out of sequence of pdf's */
boolean debug = false; /* so it does not save the generated parameters. */
/* this function generates features just for the trees and pdf that are not null in the HMM cart*/
pdf2par.htsMaximumLikelihoodParameterGeneration(um, htsData);
// (2) include the predicted values in applicableElements (as it is done in Model)
boolean voiced[] = pdf2par.getVoicedArray();
int numVoiced = 0;
// make sure that the number of applicable elements is the same as the predicted number of
// elements
assert applyToElements.size() == um.getNumModel();
float f0;
String formattedTargetValue;
int t = 0;
for (int i = 0;
i < applyToElements.size();
i++) { // this will be the same as the utterance model set
m = um.getUttModel(i);
int k = 1;
int numVoicedInModel = m.getNumVoiced();
formattedTargetValue = "";
// System.out.format("phone = %s dur_in_frames=%d num_voiced_frames=%d : ",
// m.getPhoneName(), m.getTotalDur(), numVoicedInModel);
for (int mstate = 0; mstate < cart.getNumStates(); mstate++) {
for (int frame = 0; frame < m.getDur(mstate); frame++) {
if (voiced[
t++]) { // numVoiced and t are not the same because voiced values can be true or
// false, numVoiced count just the voiced
f0 = (float) Math.exp(pdf2par.getlf0Pst().getPar(numVoiced++, 0));
formattedTargetValue +=
"("
+ Integer.toString((int) ((k * 100.0) / numVoicedInModel))
+ ","
+ Integer.toString((int) f0)
+ ")";
k++;
}
}
}
Element element = applyToElements.get(i);
// "evaluate" pseudo XPath syntax:
// TODO this needs to be extended to take into account targetAttributeNames like "foo/@bar",
// which would add the
// bar attribute to the foo child of this element, creating the child if not already
// present...
if (f0AttributeName.startsWith("@")) {
f0AttributeName = f0AttributeName.replaceFirst("@", "");
}
// format targetValue according to targetAttributeFormat
// String formattedTargetValue = String.format(targetAttributeFormat, targetValue);
// set the new attribute value:
// if the whole segment is unvoiced then f0 should not be fixed?
if (formattedTargetValue.length() > 0)
element.setAttribute(f0AttributeName, formattedTargetValue);
// System.out.println(formattedTargetValue);
}
// once finished re-set to null um
// um = null;
} catch (Exception e) {
throw new MaryConfigurationException("Error generating F0 out of HMMs trees and pdfs. ", e);
}
}
/**
* Create a utterance model list from feature vectors predicted from elements.
*
* @param predictFromElements elements from MaryXML from where to get feature vectors.
* @return Utterance model um containing state durations and pdfs already searched on the trees to
* generate F0.
* @throws MaryConfigurationException if error searching in HMM trees.
*/
private HTSUttModel createUttModel(List<Element> predictFromElements)
throws MaryConfigurationException {
int i, k, s, t, mstate, frame, durInFrames, durStateInFrames, numVoicedInModel;
HTSModel m;
List<Element> predictorElements = predictFromElements;
List<Target> predictorTargets = getTargets(predictorElements);
FeatureVector fv;
HTSUttModel um = new HTSUttModel();
FeatureDefinition feaDef = htsData.getFeatureDefinition();
float duration;
double diffdurOld = 0.0;
double diffdurNew = 0.0;
float f0s[] = null;
try {
// (1) Predict the values
for (i = 0; i < predictorTargets.size(); i++) {
fv = predictorTargets.get(i).getFeatureVector();
Element e = predictFromElements.get(i);
um.addUttModel(new HTSModel(cart.getNumStates()));
m = um.getUttModel(i);
/* this function also sets the phone name, the phone between - and + */
m.setPhoneName(fv.getFeatureAsString(feaDef.getFeatureIndex("phone"), feaDef));
/* Check if context-dependent gv (gv without sil) */
if (htsData.getUseContextDependentGV()) {
if (m.getPhoneName().contentEquals("_")) m.setGvSwitch(false);
}
/* increment number of models in utterance model */
um.setNumModel(um.getNumModel() + 1);
/* update number of states */
um.setNumState(um.getNumState() + cart.getNumStates());
// get the duration from the element
duration = Integer.parseInt(e.getAttribute("d")) * 0.001f; // in sec.
// distribute the duration (in frames) among the five states, here it is done the same
// amount for each state
durInFrames = (int) (duration / fperiodsec);
durStateInFrames = (int) (durInFrames / cart.getNumStates());
m.setTotalDur(0); // reset to set new value according to duration
for (s = 0; s < cart.getNumStates(); s++) {
m.setDur(s, durStateInFrames);
m.setTotalDur(m.getTotalDur() + m.getDur(s));
}
um.setTotalFrame(um.getTotalFrame() + m.getTotalDur());
System.out.format(
"createUttModel: duration=%.3f sec. durInFrames=%d durStateInFrames=%d m.getTotalDur()=%d\n",
duration, durInFrames, durStateInFrames, m.getTotalDur());
/* Find pdf for LF0, this function sets the pdf for each state.
* and determines, according to the HMM models, whether the states are voiced or unvoiced, (it can be possible that some states are voiced
* and some unvoiced).*/
cart.searchLf0InCartTree(m, fv, feaDef, htsData.getUV());
for (mstate = 0; mstate < cart.getNumStates(); mstate++) {
for (frame = 0; frame < m.getDur(mstate); frame++)
if (m.getVoiced(mstate)) um.setLf0Frame(um.getLf0Frame() + 1);
}
}
return um;
} catch (Exception e) {
throw new MaryConfigurationException(
"Error searching in tree when creating utterance model. ", e);
}
}
/**
* Apply the HMM to a Target to get its predicted value, this method is not used in HMMModel.
*
* @throws RuntimeException if this method is called.
*/
@Override
protected float evaluate(Target target) {
throw new RuntimeException("This method should never be called");
}
}
/**
* The unit database of a voice
*
* @author Marc Schröder
*/
public class UnitDatabase {
protected TargetCostFunction targetCostFunction;
protected JoinCostFunction joinCostFunction;
protected StatisticalCostFunction sCostFunction = null;
protected UnitFileReader unitReader;
protected int numUnits;
protected CART preselectionCART;
protected TimelineReader audioTimeline;
protected TimelineReader basenameTimeline;
protected int backtrace;
protected Logger logger = MaryUtils.getLogger("UnitDatabase");
public UnitDatabase() {}
public void load(
TargetCostFunction aTargetCostFunction,
JoinCostFunction aJoinCostFunction,
UnitFileReader aUnitReader,
CART aPreselectionCART,
TimelineReader anAudioTimeline,
TimelineReader aBasenameTimeline,
int backtraceLeafSize) {
this.targetCostFunction = aTargetCostFunction;
this.joinCostFunction = aJoinCostFunction;
this.unitReader = aUnitReader;
this.numUnits = (unitReader != null ? unitReader.getNumberOfUnits() : 0);
this.preselectionCART = aPreselectionCART;
this.audioTimeline = anAudioTimeline;
this.basenameTimeline = aBasenameTimeline;
this.backtrace = backtraceLeafSize;
}
public void load(
TargetCostFunction aTargetCostFunction,
JoinCostFunction aJoinCostFunction,
StatisticalCostFunction asCostFunction,
UnitFileReader aUnitReader,
CART aPreselectionCART,
TimelineReader anAudioTimeline,
TimelineReader aBasenameTimeline,
int backtraceLeafSize) {
this.targetCostFunction = aTargetCostFunction;
this.joinCostFunction = aJoinCostFunction;
this.sCostFunction = asCostFunction;
this.unitReader = aUnitReader;
this.numUnits = (unitReader != null ? unitReader.getNumberOfUnits() : 0);
this.preselectionCART = aPreselectionCART;
this.audioTimeline = anAudioTimeline;
this.basenameTimeline = aBasenameTimeline;
this.backtrace = backtraceLeafSize;
}
public TargetCostFunction getTargetCostFunction() {
return targetCostFunction;
}
public JoinCostFunction getJoinCostFunction() {
return joinCostFunction;
}
public UnitFileReader getUnitFileReader() {
return unitReader;
}
public TimelineReader getAudioTimeline() {
return audioTimeline;
}
public StatisticalCostFunction getSCostFunction() {
return sCostFunction;
}
/**
* Preselect a set of candidates that could be used to realise the given target.
*
* @param target a Target object representing an optimal unit
* @return an <span style="color:red;">unsorted</span> ArrayList of ViterbiCandidates, each
* containing the (same) target and a (different) Unit object
*/
public List<ViterbiCandidate> getCandidates(Target target) {
// BEGIN blacklisting
// The point of this is to get the value of the "blacklist" attribute in the first child element
// of the MaryXML
// and store it in the blacklist String variable.
// This code seems rather inelegant; perhaps there is a better way to access the MaryXML from
// this method?
String blacklist = "";
String unitBasename =
"This must never be null or the empty string!"; // otherwise candidate selection fails!
Element targetElement = target.getMaryxmlElement();
blacklist = DomUtils.getAttributeFromClosestAncestorOfAnyKind(targetElement, "blacklist");
// END blacklisting
// logger.debug("Looking for candidates in cart "+target.getName());
// get the cart tree and extract the candidates
int[] clist = (int[]) preselectionCART.interpret(target, backtrace);
logger.debug("For target " + target + ", selected " + clist.length + " units");
// Now, clist is an array of unit indexes.
List<ViterbiCandidate> candidates = new ArrayList<ViterbiCandidate>();
for (int i = 0; i < clist.length; i++) {
// The target is the same for all these candidates in the queue
// remember the actual unit:
Unit unit = unitReader.getUnit(clist[i]);
candidates.add(new ViterbiCandidate(target, unit, targetCostFunction));
}
// Blacklisting without crazy performance drop:
// just remove candidates again if their basenames are blacklisted
java.util.Iterator<ViterbiCandidate> candIt = candidates.iterator();
while (candIt.hasNext()) {
ViterbiCandidate candidate = candIt.next();
unitBasename = getFilename(candidate.getUnit());
if (blacklist.contains(unitBasename)) {
candIt.remove();
}
}
return candidates;
}
/**
* For debugging, return the basename of the original audio file from which the unit is coming, as
* well as the start time in that file.
*
* @param unit unit
* @return a String containing basename followed by a space and the unit's start time, in seconds,
* from the beginning of the file. If no basenameTimeline was specified for this voice,
* returns the string "unknown origin".
*/
public String getFilenameAndTime(Unit unit) {
if (basenameTimeline == null) return "unknown origin";
long[] offset = new long[1];
try {
Datagram[] datagrams =
basenameTimeline.getDatagrams(unit.startTime, 1, unitReader.getSampleRate(), offset);
Datagram filenameData = datagrams[0];
float time = (float) offset[0] / basenameTimeline.getSampleRate();
String filename = new String(filenameData.getData(), "UTF-8");
return filename + " " + time;
} catch (Exception e) {
logger.warn(
"Problem getting filename and time for unit " + unit.index + " at time " + unit.startTime,
e);
return "unknown origin";
}
}
/**
* For debugging, return the basename of the original audio file from which the unit is coming.
*
* @param unit unit
* @return a String containing basename. If no basenameTimeline was specified for this voice,
* returns the string "unknown origin".
*/
public String getFilename(Unit unit) {
// if (basenameTimeline == null) return "unknown origin";
try {
Datagram filenameData = basenameTimeline.getDatagram(unit.startTime);
String filename = new String(filenameData.getData(), "UTF-8");
return filename;
} catch (Exception e) {
logger.warn("Problem getting filename for unit " + unit.index, e);
return "unknown origin";
}
}
}
public void computeFeaturesFor(String basename) throws IOException, Exception {
String text;
Locale localVoice;
localVoice = MaryUtils.string2locale(locale);
// First, test if there is a corresponding .rawmaryxml file in textdir:
File rawmaryxmlFile =
new File(db.getProp(db.MARYXMLDIR) + basename + db.getProp(db.MARYXMLEXT));
if (rawmaryxmlFile.exists()) {
text = FileUtils.getFileAsString(rawmaryxmlFile, "UTF-8");
} else {
text =
getMaryXMLHeaderWithInitialBoundary(locale)
+ FileUtils.getFileAsString(
new File(db.getProp(db.TEXTDIR) + basename + db.getProp(db.TEXTEXT)), "UTF-8")
+ "</maryxml>";
}
File pfeatFile = new File(unitfeatureDir, basename + featsExt);
OutputStream os = new BufferedOutputStream(new FileOutputStream(pfeatFile));
MaryClient maryClient = getMaryClient();
/*Vector voices = maryClient.getVoices(localVoice);
MaryClient.Voice defaultVoice = (MaryClient.Voice) voices.firstElement();
String voiceName = defaultVoice.name();*/
// maryClient.process(text, maryInputType, maryOutputType, null, null, os);
maryClient.process(text, maryInputType, maryOutputType, locale, null, "slt-arctic", os);
// maryClient.process(text, maryInputType, maryOutputType, null, "slt-arctic", os, timeout);
// maryClient.getOutputDataTypes().size()
// MaryData result = new MaryData(os);
os.flush();
os.close();
// System.out.println(" TO STRING: "+new FileReader(pfeatFile).toString());
// BufferedReader bfr = new BufferedReader(new FileReader(pfeatFile));
String line;
MaryData d = new MaryData(MaryDataType.get("PHONEMISED_EN"), Locale.US);
// d.readFrom(new ByteArrayInputStream(os.toByteArray()));
d.readFrom(new FileReader(pfeatFile));
// MaryData d = new MaryData(pfeatFile);
Document doc = d.getDocument();
// Document acoustparams = d.getDocument();
// NodeIterator it = ((DocumentTraversal)acoustparams).createNodeIterator(acoustparams,
// NodeFilter.SHOW_ELEMENT,new NameNodeFilter(new String[]{MaryXML.TOKEN,
// MaryXML.BOUNDARY}),false);
NodeIterator it =
((DocumentTraversal) doc)
.createNodeIterator(
doc, NodeFilter.SHOW_ELEMENT, new NameNodeFilter(MaryXML.TOKEN), false);
Element t = null;
while ((t = (Element) it.nextNode()) != null) {
if (t.hasAttribute("g2p_method")) {
String g2p = t.getAttribute("g2p_method");
String nodeText = t.getTextContent().trim();
if (g2p.equals("rules")) { // && nodeText.equals("!")){
System.out.print(basename + " ----> " + nodeText);
if (bnl.contains(basename)) bnl.remove(basename);
System.out.println(" SO removing basename: " + basename);
}
// System.out.println("G2P:"+t.getAttribute("g2p_method"));
// System.out.println("Text:"+t.getTextContent());
}
}
/*while((line =bfr.readLine()) != null){
//boolean b = m.matches();
if(Pattern.matches("rules", line))
System.out.println(basename + " LINE ---> " + line);
}*/
// System.out.println(" TO STRING: "+line);
}
/**
* Parameter generation out of trained HMMs.
*
* <p>Java port and extension of HTS engine API version 1.04 Extension: mixed excitation
*
* @author Marcela Charfuelan
*/
public class HTSParameterGeneration {
public static final double INFTY = ((double) 1.0e+38);
public static final double INFTY2 = ((double) 1.0e+19);
public static final double INVINF = ((double) 1.0e-38);
public static final double INVINF2 = ((double) 1.0e-19);
private HTSPStream mcepPst = null;
private HTSPStream strPst = null;
private HTSPStream magPst = null;
private HTSPStream lf0Pst = null;
private boolean voiced[];
private int totalUttFrame; // total number of frames in a mcep, str or mag Pst
private int totalLf0Frame; // total number of f0 voiced frames in a lf0 Pst
private Logger logger = MaryUtils.getLogger("ParameterGeneration");
public HTSPStream getMcepPst() {
return mcepPst;
}
public void setMcepPst(HTSPStream var) {
mcepPst = var;
};
public HTSPStream getStrPst() {
return strPst;
}
public void setStrPst(HTSPStream var) {
strPst = var;
};
public HTSPStream getMagPst() {
return magPst;
}
public void setMagPst(HTSPStream var) {
magPst = var;
};
public HTSPStream getlf0Pst() {
return lf0Pst;
}
public void setlf0Pst(HTSPStream var) {
lf0Pst = var;
};
public boolean[] getVoicedArray() {
return voiced;
}
public void setVoicedArray(boolean[] var) {
voiced = var;
} // only used in HTSEngineTest
/* Inverse of a given double */
/* We actually need the inverse of the matrix of covariance, but since this matrix */
/* is a diagonal matrix, then we just need to calculate the inverse of each of the */
/* numbers in the diagonal. */
public static double finv(double x) {
if (x >= INFTY2) return 0.0;
if (x <= -INFTY2) return 0.0;
if (x <= INVINF2 && x >= 0) return INFTY;
if (x >= -INVINF2 && x < 0) return -INFTY;
return 1.0 / x;
}
/**
* HTS maximum likelihood parameter generation
*
* @param um : utterance model sequence after processing Mary context features
* @param htsData : HMM pdfs model set.
* @throws Exception Exception
*/
public void htsMaximumLikelihoodParameterGeneration(HTSUttModel um, final HMMData htsData)
throws Exception {
CartTreeSet ms = htsData.getCartTreeSet();
/* Initialisation of PStream objects */
/* Initialise Parameter generation using UttModel um and Modelset ms */
/* initialise PStream objects for all the parameters that are going to be generated: */
/* mceppst, strpst, magpst, lf0pst */
/* Here i should pass the window files to initialise the dynamic windows dw */
/* for the moment the dw are all the same and hard-coded */
if (htsData.getPdfMgcStream() != null)
mcepPst =
new HTSPStream(
ms.getMcepVsize(),
um.getTotalFrame(),
HMMData.FeatureType.MGC,
htsData.getMaxMgcGvIter());
/* for lf0 count just the number of lf0frames that are voiced or non-zero */
if (htsData.getPdfLf0Stream() != null)
lf0Pst =
new HTSPStream(
ms.getLf0Stream(),
um.getLf0Frame(),
HMMData.FeatureType.LF0,
htsData.getMaxLf0GvIter());
/* The following are optional in case of generating mixed excitation */
if (htsData.getPdfStrStream() != null)
strPst =
new HTSPStream(
ms.getStrVsize(),
um.getTotalFrame(),
HMMData.FeatureType.STR,
htsData.getMaxStrGvIter());
if (htsData.getPdfMagStream() != null)
magPst =
new HTSPStream(
ms.getMagVsize(),
um.getTotalFrame(),
HMMData.FeatureType.MAG,
htsData.getMaxMagGvIter());
int lf0Frame = 0; // counts voiced frames
int uttFrame = 0; // counts all frames
voiced = new boolean[um.getTotalFrame()];
// local variables for faster access
int msNumStates = ms.getNumStates();
int totalFrames = um.getTotalFrame();
for (int i = 0; i < um.getNumUttModel(); i++) {
HTSModel m = um.getUttModel(i);
int numVoicedInModel = 0;
for (int state = 0; state < msNumStates; state++) {
int dur = m.getDur(state);
Arrays.fill(voiced, uttFrame, uttFrame += dur, m.getVoiced(state));
if (m.getVoiced(state)) lf0Frame += dur;
}
}
/* mcepframe and lf0frame are used in the original code to initialise the T field */
/* in each pst, but here the pst are already initialised .... */
logger.debug("utteranceFrame=" + uttFrame + " lf0frame=" + lf0Frame);
// Step 1: initialize fields in the parameter streams
uttFrame = 0;
lf0Frame = 0;
/* copy pdfs */
for (int i = 0; i < um.getNumUttModel(); i++) {
HTSModel m = um.getUttModel(i);
boolean gvSwitch = m.getGvSwitch();
for (int state = 0; state < msNumStates; state++) {
for (int frame = 0; frame < m.getDur(state); frame++) {
/* copy pdfs for mcep */
if (mcepPst != null) {
mcepPst.setMseq(uttFrame, m.getMean(FeatureType.MGC, state));
mcepPst.setVseq(uttFrame, m.getVariance(FeatureType.MGC, state));
if (!gvSwitch) mcepPst.setGvSwitch(uttFrame, false);
}
/* copy pdf for str */
if (strPst != null) {
strPst.setMseq(uttFrame, m.getMean(FeatureType.STR, state));
strPst.setVseq(uttFrame, m.getVariance(FeatureType.STR, state));
if (!gvSwitch) strPst.setGvSwitch(uttFrame, false);
}
/* copy pdf for mag */
if (magPst != null) {
magPst.setMseq(uttFrame, m.getMean(FeatureType.MAG, state));
magPst.setVseq(uttFrame, m.getVariance(FeatureType.MAG, state));
if (!gvSwitch) magPst.setGvSwitch(uttFrame, false);
}
/* copy pdfs for lf0 */
if (lf0Pst != null && !htsData.getUseAcousticModels()) {
for (int k = 0; k < ms.getLf0Stream(); k++) {
boolean nobound = true;
/* check if current frame is voiced/unvoiced boundary or not */
for (int n = lf0Pst.getDWLeftBoundary(k); n <= lf0Pst.getDWRightBoundary(k); n++)
if ((uttFrame + n) <= 0 || totalFrames <= (uttFrame + n)) nobound = false;
else nobound = (nobound && voiced[uttFrame + n]);
/* copy pdfs */
if (voiced[uttFrame]) {
lf0Pst.setMseq(lf0Frame, k, m.getLf0Mean(state, k));
if (nobound || k == 0)
lf0Pst.setIvseq(lf0Frame, k, finv(m.getLf0Variance(state, k)));
else
/* the variances for dynamic features are set to inf on v/uv boundary */
lf0Pst.setIvseq(lf0Frame, k, 0.0);
}
}
}
if (voiced[uttFrame]) {
if (!gvSwitch) lf0Pst.setGvSwitch(lf0Frame, false);
lf0Frame++;
}
uttFrame++;
} /* for each frame in this state */
} /* for each state in this model */
} /* for each model in this utterance */
GVModelSet gvms = htsData.getGVModelSet();
// Step 2: set dynamic features to infinity on the borders for MGC/STR/MAG
if (mcepPst != null) mcepPst.fixDynFeatOnBoundaries();
if (strPst != null) strPst.fixDynFeatOnBoundaries();
if (magPst != null) magPst.fixDynFeatOnBoundaries();
// Step 3: optimize individual parameter streams
/* parameter generation for mcep */
if (mcepPst != null) {
logger.info("Parameter generation for MGC: ");
if (htsData.getUseGV() && (htsData.getPdfMgcGVStream() != null))
mcepPst.setGvMeanVar(gvms.getGVmeanMgc(), gvms.getGVcovInvMgc());
mcepPst.mlpg(htsData, htsData.getUseGV());
}
// parameter generation for lf0 */
if (htsData.getUseAcousticModels()) loadMaryXmlF0(um, htsData);
else if (lf0Pst != null) {
logger.info("Parameter generation for LF0: ");
if (htsData.getUseGV() && (htsData.getPdfLf0GVStream() != null))
lf0Pst.setGvMeanVar(gvms.getGVmeanLf0(), gvms.getGVcovInvLf0());
lf0Pst.mlpg(htsData, htsData.getUseGV());
// here we need set realisedF0
setRealisedF0(lf0Pst, um, msNumStates);
}
/* parameter generation for str */
boolean useGV = false;
if (strPst != null) {
logger.debug("Parameter generation for STR ");
if (htsData.getUseGV() && (htsData.getPdfStrGVStream() != null)) {
useGV = true;
strPst.setGvMeanVar(gvms.getGVmeanStr(), gvms.getGVcovInvStr());
}
strPst.mlpg(htsData, useGV);
}
/* parameter generation for mag */
useGV = false;
if (magPst != null) {
logger.info("Parameter generation for MAG ");
if (htsData.getUseGV() && (htsData.getPdfMagGVStream() != null)) {
useGV = true;
magPst.setGvMeanVar(gvms.getGVmeanMag(), gvms.getGVcovInvMag());
}
magPst.mlpg(htsData, useGV);
}
} /* method htsMaximumLikelihoodParameterGeneration */
/* Save generated parameters in a binary file */
public void saveParamMaryFormat(String fileName, HTSPStream par, HMMData.FeatureType type) {
int t, m, i;
double ws = 0.025; /* window size in seconds */
double ss = 0.005; /* skip size in seconds */
int fs = 16000; /* sampling rate */
try {
if (type == HMMData.FeatureType.LF0) {
fileName += ".ptc";
/*
* DataOutputStream data_out = new DataOutputStream (new FileOutputStream (fileName));
* data_out.writeFloat((float)(ws*fs)); data_out.writeFloat((float)(ss*fs)); data_out.writeFloat((float)fs);
* data_out.writeFloat(voiced.length);
*
* i=0; for(t=0; t<voiced.length; t++){ // here par.getT are just the voiced!!! so the actual length of frames can
* be taken from the voiced array if( voiced[t] ){ data_out.writeFloat((float)Math.exp(par.getPar(i,0))); i++;
* }System.out.println("GEN f0s[" + t + "]=" + Math.exp(lf0Pst.getPar(i,0))); else
* data_out.writeFloat((float)0.0); } data_out.close();
*/
i = 0;
double f0s[] = new double[voiced.length];
// System.out.println("voiced.length=" + voiced.length);
for (t = 0;
t < voiced.length;
t++) { // here par.getT are just the voiced!!! so the actual length of frames can
// be taken from the voiced array
if (voiced[t]) {
f0s[t] = Math.exp(par.getPar(i, 0));
i++;
} else f0s[t] = 0.0;
System.out.println("GEN f0s[" + t + "]=" + f0s[t]);
}
/*
* i am using this function but it changes the values of sw, and ss *samplingrate+0.5??? for the HTS values
* ss=0.005 and sw=0.025 is not a problem though
*/
PitchReaderWriter.write_pitch_file(fileName, f0s, (float) (ws), (float) (ss), fs);
} else if (type == HMMData.FeatureType.MGC) {
int numfrm = par.getT();
int dimension = par.getOrder();
Mfccs mgc = new Mfccs(numfrm, dimension);
fileName += ".mfc";
for (t = 0; t < par.getT(); t++)
for (m = 0; m < par.getOrder(); m++) mgc.mfccs[t][m] = par.getPar(t, m);
mgc.params.samplingRate = fs; /* samplingRateInHz */
mgc.params.skipsize = (float) ss; /* skipSizeInSeconds */
mgc.params.winsize = (float) ws; /* windowSizeInSeconds */
mgc.writeMfccFile(fileName);
/*
* The whole set for header is in the following order: ler.writeInt(numfrm); ler.writeInt(dimension);
* ler.writeFloat(winsize); ler.writeFloat(skipsize); ler.writeInt(samplingRate);
*/
}
logger.info("saveParam in file: " + fileName);
} catch (IOException e) {
logger.info("IO exception = " + e);
}
}
/* Save generated parameters in a binary file */
public void saveParam(String fileName, HTSPStream par, HMMData.FeatureType type) {
int t, m, i;
try {
if (type == HMMData.FeatureType.LF0) {
fileName += ".f0";
DataOutputStream data_out = new DataOutputStream(new FileOutputStream(fileName));
i = 0;
for (t = 0; t < voiced.length; t++) {
/* here par.getT are just the voiced!!! */
if (voiced[t]) {
data_out.writeFloat((float) Math.exp(par.getPar(i, 0)));
i++;
} else data_out.writeFloat((float) 0.0);
}
data_out.close();
} else if (type == HMMData.FeatureType.MGC) {
fileName += ".mgc";
DataOutputStream data_out = new DataOutputStream(new FileOutputStream(fileName));
for (t = 0; t < par.getT(); t++)
for (m = 0; m < par.getOrder(); m++) data_out.writeFloat((float) par.getPar(t, m));
data_out.close();
}
logger.info("saveParam in file: " + fileName);
} catch (IOException e) {
logger.info("IO exception = " + e);
}
}
private void loadMaryXmlF0(HTSUttModel um, HMMData htsData) throws Exception {
logger.info("Using f0 from maryXML acoustparams");
int i, n, numVoiced;
HTSModel m;
double[] dval;
double lastF0 = 0.0;
numVoiced = 0;
Vector<Double> f0Vector = new Vector<Double>();
for (i = 0; i < um.getNumUttModel(); i++) {
m = um.getUttModel(i);
// System.out.format("\nmodel=%s totalDur=%d numVoicedFrames=%d F0=%s\n", m.getPhoneName(),
// m.getTotalDur(),
// m.getNumVoiced(), m.getMaryXmlF0());
// get contour for this model if voiced frames and maryXml has f0 values
dval = getContourSegment(m.getMaryXmlF0(), m.getNumVoiced());
// accumulate the values
for (n = 0; n < dval.length; n++) f0Vector.add(dval[n]);
}
// interpolate values if necessary
interpolateSegments(f0Vector);
// create a new Lf0Pst with the values from maryXML
HTSPStream newLf0Pst =
new HTSPStream(3, f0Vector.size(), HMMData.FeatureType.LF0, htsData.getMaxLf0GvIter());
for (n = 0; n < f0Vector.size(); n++) newLf0Pst.setPar(n, 0, Math.log(f0Vector.get(n)));
setlf0Pst(newLf0Pst);
}
private double[] getContourSegment(String maryXmlF0, int numVoiced) throws Exception {
int i, t = 0, k = 0, f = 0; // f is number of f0 in xml string
// just fill the values in approx. position
double[] f0Vector = new double[numVoiced];
int index[] = new int[2];
double value[] = new double[2];
int key, n, interval;
double valF0, lastValF0;
if (maryXmlF0 != null) {
Pattern p = Pattern.compile("(\\d+,\\d+)");
Matcher xml = p.matcher(maryXmlF0);
SortedMap<Integer, Double> f0Map = new TreeMap<Integer, Double>();
int numF0s = 0;
while (xml.find()) {
String[] f0Values = (xml.group().trim()).split(",");
f0Map.put(new Integer(f0Values[0]), new Double(f0Values[1]));
numF0s++;
}
Set<Map.Entry<Integer, Double>> s = f0Map.entrySet();
Iterator<Map.Entry<Integer, Double>> if0 = s.iterator();
if (numF0s == numVoiced) {
t = 0;
while (if0.hasNext() && t < numVoiced) {
Map.Entry<Integer, Double> mf0 = if0.next();
key = (Integer) mf0.getKey();
valF0 = (Double) mf0.getValue();
f0Vector[t++] = valF0;
}
} else {
if (numF0s < numVoiced) {
for (i = 0; i < numVoiced; i++)
// then just some values will be filled, so the other must be 0
f0Vector[i] = 0.0;
}
while (if0.hasNext() && t < numVoiced) {
Map.Entry<Integer, Double> mf0 = if0.next();
key = (Integer) mf0.getKey();
valF0 = (Double) mf0.getValue();
if (key == 0) n = 0;
else if (key == 100) n = numVoiced - 1;
else n = (int) ((numVoiced * key) / 100.0);
if (n >= 0 && n < numVoiced) f0Vector[n] = valF0;
} // while(if0.hasNext())
} // numF0s == numVoiced
} // if maryXML != null
// for(i=0; i<numVoiced; i++) // then just some values will be filled, so the other must be 0
// System.out.format("%.1f ", f0Vector[i]);
// System.out.println();
return f0Vector;
}
private void interpolateSegments(Vector<Double> f0) {
int i, n, interval;
double slope;
// check where there are zeros and interpolate
int[] index = new int[2];
double[] value = new double[2];
index[0] = 0;
value[0] = 0.0;
for (i = 0; i < f0.size(); i++) {
if (f0.get(i) > 0.0) {
index[1] = i;
value[1] = f0.get(i);
interval = index[1] - index[0];
if (interval > 1) {
// System.out.format("Interval to interpolate index[0]=%d
// index[1]=%d\n",index[0],index[1]);
slope = ((value[1] - value[0]) / interval);
for (n = index[0]; n < index[1]; n++) {
double newVal = (slope * (n - index[0])) + value[0];
f0.set(n, newVal);
// System.out.format(" n=%d value:%.1f\n",n,newVal);
}
}
index[0] = index[1];
value[0] = value[1];
}
}
}
private void setRealisedF0(HTSPStream lf0Pst, HTSUttModel um, int numStates) {
int t = 0;
int vt = 0;
for (int i = 0; i < um.getNumUttModel(); i++) {
HTSModel m = um.getUttModel(i);
int numVoicedInModel = m.getNumVoiced();
String formattedF0 = "";
int k = 1;
for (int state = 0; state < numStates; state++) {
for (int frame = 0; frame < m.getDur(state); frame++) {
if (voiced[t++]) {
float f0 = (float) Math.exp(lf0Pst.getPar(vt++, 0));
formattedF0 +=
"("
+ Integer.toString((int) ((k * 100.0) / numVoicedInModel))
+ ","
+ Integer.toString((int) f0)
+ ")";
k++;
}
} // for unvoiced frame
} // for state
if (!formattedF0.contentEquals("")) {
m.setMaryXmlF0(formattedF0);
// m.setUnit_f0ArrayStr(formattedF0);
// System.out.println("ph=" + m.getPhoneName() + " " + formattedF0);
}
} // for model in utterance model list
}
} /* class ParameterGeneration */
/**
* IO functions for directed graphs in Mary format
*
* @author Marcela Charfuelan, Marc Schröder
*/
public class DirectedGraphWriter {
protected Logger logger = MaryUtils.getLogger(this.getClass().getName());
/**
* Dump the graph in Mary format
*
* @param graph graph
* @param destFile the destination file
* @throws IOException IOException
*/
public void saveGraph(DirectedGraph graph, String destFile) throws IOException {
if (graph == null) throw new NullPointerException("Cannot dump null graph");
if (destFile == null) throw new NullPointerException("No destination file");
logger.debug("Dumping directed graph in Mary format to " + destFile + " ...");
// Open the destination file and output the header
DataOutputStream out =
new DataOutputStream(new BufferedOutputStream(new FileOutputStream(destFile)));
// create new CART-header and write it to output file
MaryHeader hdr = new MaryHeader(MaryHeader.DIRECTED_GRAPH);
hdr.writeTo(out);
Properties props = graph.getProperties();
if (props == null) {
out.writeShort(0);
} else {
ByteArrayOutputStream baos = new ByteArrayOutputStream();
props.store(baos, null);
byte[] propData = baos.toByteArray();
out.writeShort(propData.length);
out.write(propData);
}
// feature definition
graph.getFeatureDefinition().writeBinaryTo(out);
// dump graph
dumpBinary(graph, out);
// finish
out.close();
logger.debug(" ... done\n");
}
public void toTextOut(DirectedGraph graph, PrintWriter pw) throws IOException {
try {
int numLeafNodes = setUniqueLeafNodeIds(graph);
int numDecNodes = setUniqueDecisionNodeIds(graph);
int numGraphNodes = setUniqueDirectedGraphNodeIds(graph);
pw.println(
"Num decision nodes= "
+ numDecNodes
+ " Num leaf nodes= "
+ numLeafNodes
+ " Num directed graph nodes= "
+ numGraphNodes);
printDecisionNodes(graph, null, pw);
pw.println("\n----------------\n");
printLeafNodes(graph, null, pw);
pw.println("\n----------------\n");
printDirectedGraphNodes(graph, null, pw);
pw.flush();
pw.close();
} catch (IOException ioe) {
IOException newIOE = new IOException("Error dumping graph to standard output");
newIOE.initCause(ioe);
throw newIOE;
}
}
/**
* Assign unique ids to leaf nodes.
*
* @param graph
* @return the number of different leaf nodes
*/
private int setUniqueLeafNodeIds(DirectedGraph graph) {
int i = 0;
for (LeafNode l : graph.getLeafNodes()) {
l.setUniqueLeafId(++i);
}
return i;
}
/**
* Assign unique ids to decision nodes.
*
* @param graph
* @return the number of different decision nodes
*/
private int setUniqueDecisionNodeIds(DirectedGraph graph) {
int i = 0;
for (DecisionNode d : graph.getDecisionNodes()) {
d.setUniqueDecisionNodeId(++i);
}
return i;
}
/**
* Assign unique ids to directed graph nodes.
*
* @param graph
* @return the number of different directed graph nodes
*/
private int setUniqueDirectedGraphNodeIds(DirectedGraph graph) {
int i = 0;
for (DirectedGraphNode g : graph.getDirectedGraphNodes()) {
g.setUniqueGraphNodeID(++i);
}
return i;
}
private void dumpBinary(DirectedGraph graph, DataOutput os) throws IOException {
try {
int numLeafNodes = setUniqueLeafNodeIds(graph);
int numDecNodes = setUniqueDecisionNodeIds(graph);
int numGraphNodes = setUniqueDirectedGraphNodeIds(graph);
int maxNum = 1 << 30;
if (numLeafNodes > maxNum || numDecNodes > maxNum || numGraphNodes > maxNum) {
throw new UnsupportedOperationException(
"Cannot write more than " + maxNum + " nodes of one type in this format");
}
// write the number of decision nodes
os.writeInt(numDecNodes);
printDecisionNodes(graph, os, null);
// write the number of leaves.
os.writeInt(numLeafNodes);
printLeafNodes(graph, os, null);
// write the number of directed graph nodes
os.writeInt(numGraphNodes);
printDirectedGraphNodes(graph, os, null);
} catch (IOException ioe) {
IOException newIOE = new IOException("Error dumping CART to output stream");
newIOE.initCause(ioe);
throw newIOE;
}
}
private void printDecisionNodes(DirectedGraph graph, DataOutput out, PrintWriter pw)
throws IOException {
for (DecisionNode decNode : graph.getDecisionNodes()) {
int id = decNode.getUniqueDecisionNodeId();
String nodeDefinition = decNode.getNodeDefinition();
int featureIndex = decNode.getFeatureIndex();
DecisionNode.Type nodeType = decNode.getDecisionNodeType();
if (out != null) {
// dump in binary form to output
out.writeInt(featureIndex);
out.writeInt(nodeType.ordinal());
// Now, questionValue, which depends on nodeType
switch (nodeType) {
case BinaryByteDecisionNode:
out.writeInt(((BinaryByteDecisionNode) decNode).getCriterionValueAsByte());
assert decNode.getNumberOfDaugthers() == 2;
break;
case BinaryShortDecisionNode:
out.writeInt(((BinaryShortDecisionNode) decNode).getCriterionValueAsShort());
assert decNode.getNumberOfDaugthers() == 2;
break;
case BinaryFloatDecisionNode:
out.writeFloat(((BinaryFloatDecisionNode) decNode).getCriterionValueAsFloat());
assert decNode.getNumberOfDaugthers() == 2;
break;
case ByteDecisionNode:
case ShortDecisionNode:
out.writeInt(decNode.getNumberOfDaugthers());
}
// The child nodes
for (int i = 0, n = decNode.getNumberOfDaugthers(); i < n; i++) {
Node daughter = decNode.getDaughter(i);
if (daughter == null) {
out.writeInt(0);
} else if (daughter.isDecisionNode()) {
int daughterID = ((DecisionNode) daughter).getUniqueDecisionNodeId();
// Mark as decision node:
daughterID |= DirectedGraphReader.DECISIONNODE << 30;
out.writeInt(daughterID);
} else if (daughter.isLeafNode()) {
int daughterID = ((LeafNode) daughter).getUniqueLeafId();
// Mark as leaf node:
if (daughterID != 0) daughterID |= DirectedGraphReader.LEAFNODE << 30;
out.writeInt(daughterID);
} else if (daughter.isDirectedGraphNode()) {
int daughterID = ((DirectedGraphNode) daughter).getUniqueGraphNodeID();
// Mark as directed graph node:
if (daughterID != 0) daughterID |= DirectedGraphReader.DIRECTEDGRAPHNODE << 30;
out.writeInt(daughterID);
}
}
}
if (pw != null) {
// dump to print writer
StringBuilder strNode = new StringBuilder("-" + id + " " + nodeDefinition);
for (int i = 0, n = decNode.getNumberOfDaugthers(); i < n; i++) {
strNode.append(" ");
Node daughter = decNode.getDaughter(i);
if (daughter == null) {
strNode.append("0");
} else if (daughter.isDecisionNode()) {
int daughterID = ((DecisionNode) daughter).getUniqueDecisionNodeId();
strNode.append("-").append(daughterID);
out.writeInt(daughterID);
} else if (daughter.isLeafNode()) {
int daughterID = ((LeafNode) daughter).getUniqueLeafId();
if (daughterID == 0) strNode.append("0");
else strNode.append("id").append(daughterID);
} else if (daughter.isDirectedGraphNode()) {
int daughterID = ((DirectedGraphNode) daughter).getUniqueGraphNodeID();
if (daughterID == 0) strNode.append("0");
else strNode.append("DGN").append(daughterID);
}
}
pw.println(strNode.toString());
}
}
}
private void printLeafNodes(DirectedGraph graph, DataOutput out, PrintWriter pw)
throws IOException {
for (LeafNode leaf : graph.getLeafNodes()) {
if (leaf.getUniqueLeafId() == 0) // empty leaf, do not write
continue;
LeafType leafType = leaf.getLeafNodeType();
if (leafType == LeafType.FeatureVectorLeafNode) {
leafType = LeafType.IntArrayLeafNode;
// save feature vector leaf nodes as int array leaf nodes
}
if (out != null) {
// Leaf node type
out.writeInt(leafType.ordinal());
}
if (pw != null) {
pw.print("id" + leaf.getUniqueLeafId() + " " + leafType);
}
switch (leaf.getLeafNodeType()) {
case IntArrayLeafNode:
int data[] = ((IntArrayLeafNode) leaf).getIntData();
// Number of data points following:
if (out != null) out.writeInt(data.length);
if (pw != null) pw.print(" " + data.length);
// for each index, write the index
for (int i = 0; i < data.length; i++) {
if (out != null) out.writeInt(data[i]);
if (pw != null) pw.print(" " + data[i]);
}
break;
case FloatLeafNode:
float stddev = ((FloatLeafNode) leaf).getStDeviation();
float mean = ((FloatLeafNode) leaf).getMean();
if (out != null) {
out.writeFloat(stddev);
out.writeFloat(mean);
}
if (pw != null) {
pw.print(" 1 " + stddev + " " + mean);
}
break;
case IntAndFloatArrayLeafNode:
case StringAndFloatLeafNode:
int data1[] = ((IntAndFloatArrayLeafNode) leaf).getIntData();
float floats[] = ((IntAndFloatArrayLeafNode) leaf).getFloatData();
// Number of data points following:
if (out != null) out.writeInt(data1.length);
if (pw != null) pw.print(" " + data1.length);
// for each index, write the index and then its float
for (int i = 0; i < data1.length; i++) {
if (out != null) {
out.writeInt(data1[i]);
out.writeFloat(floats[i]);
}
if (pw != null) pw.print(" " + data1[i] + " " + floats[i]);
}
break;
case FeatureVectorLeafNode:
FeatureVector fv[] = ((FeatureVectorLeafNode) leaf).getFeatureVectors();
// Number of data points following:
if (out != null) out.writeInt(fv.length);
if (pw != null) pw.print(" " + fv.length);
// for each feature vector, write the index
for (int i = 0; i < fv.length; i++) {
if (out != null) out.writeInt(fv[i].getUnitIndex());
if (pw != null) pw.print(" " + fv[i].getUnitIndex());
}
break;
case PdfLeafNode:
throw new IllegalArgumentException("Writing of pdf leaf nodes not yet implemented");
}
if (pw != null) pw.println();
}
}
private void printDirectedGraphNodes(DirectedGraph graph, DataOutput out, PrintWriter pw)
throws IOException {
for (DirectedGraphNode g : graph.getDirectedGraphNodes()) {
int id = g.getUniqueGraphNodeID();
if (id == 0) continue; // empty node, do not write
Node leaf = g.getLeafNode();
int leafID = 0;
int leafNodeType = 0;
if (leaf != null) {
if (leaf instanceof LeafNode) {
leafID = ((LeafNode) leaf).getUniqueLeafId();
leafNodeType = DirectedGraphReader.LEAFNODE;
} else if (leaf instanceof DirectedGraphNode) {
leafID = ((DirectedGraphNode) leaf).getUniqueGraphNodeID();
leafNodeType = DirectedGraphReader.DIRECTEDGRAPHNODE;
} else {
throw new IllegalArgumentException("Unexpected leaf type: " + leaf.getClass());
}
}
DecisionNode d = g.getDecisionNode();
int decID = d != null ? d.getUniqueDecisionNodeId() : 0;
if (out != null) {
int outLeafId = leafID == 0 ? 0 : leafID | (leafNodeType << 30);
out.writeInt(outLeafId);
int outDecId = decID == 0 ? 0 : decID | (DirectedGraphReader.DECISIONNODE << 30);
out.writeInt(outDecId);
}
if (pw != null) {
pw.print("DGN" + id);
if (leafID == 0) {
pw.print(" 0");
} else if (leaf.isLeafNode()) {
pw.print(" id" + leafID);
} else {
assert leaf.isDirectedGraphNode();
pw.print(" DGN" + leafID);
}
if (decID == 0) pw.print(" 0");
else pw.print(" -" + decID);
pw.println();
}
}
}
}
public MaryHttpServer() {
logger = MaryUtils.getLogger("server");
}
/** @author steigner */
public class PhonemiseDenglish {
private static Logger logger = MaryUtils.getLogger("PhonemiseDenglish");
private String[] vowels = {"a", "e", "i", "o", "u"};
private String[] dentalPlosives = {"t", "d"};
private Hashtable<String, String> flectionToPhon = null;
private Hashtable<String, String> prefixLexicon = null;
private Hashtable<String, String> endingsAndAffixes = null;
private Hashtable<String, String> terminalVoicings = null;
private HashSet<String> endingSet = null;
private int maxEndingLength = 0;
private int maxPrefixLength = 0;
private Locale locale;
private JPhonemiser jphon = null; // building an instance of JPhonemiser for using
// lexiconLookup method and transducers from this class
private MorphologyReader mr = new MorphologyReader();
public PhonemiseDenglish(JPhonemiser jphon) throws Exception {
this.jphon = jphon;
String classpathPrefix = "/marytts/language/de/lexicon/denglish/";
this.flectionToPhon =
mr.loadInputModel(getClass().getResourceAsStream(classpathPrefix + "flectionsPhonLex.xml"));
this.prefixLexicon =
mr.loadInputModel(getClass().getResourceAsStream(classpathPrefix + "PrefixLex.xml"));
for (Iterator<String> prefixIt = prefixLexicon.keySet().iterator(); prefixIt.hasNext(); ) {
String prefix = prefixIt.next();
if (prefix.length() > this.maxPrefixLength) this.maxPrefixLength = prefix.length();
}
this.endingsAndAffixes =
mr.loadInputModel(getClass().getResourceAsStream(classpathPrefix + "germanEndings.xml"));
this.terminalVoicings =
mr.loadInputModel(
getClass().getResourceAsStream(classpathPrefix + "terminal_voicing_for_german.xml"));
String[] endingList =
getEndingsAndAffixes("flections"); // list of flection endings of a specific language
if (endingList != null) {
this.endingSet = new HashSet<String>(50); //
for (int j = 0; j < endingList.length; j++) {
if (endingList[j].length() > this.maxEndingLength) {
this.maxEndingLength = endingList[j].length();
}
this.endingSet.add(endingList[j]);
}
}
}
/**
* Method that is called from JPhonemiser - Managing all the processing
*
* @param toBePhonemised The input word
* @param allowOtherLanguage allowOtherLanguage
* @return Transcription of input word if one can be built - null otherwise
*/
public Result processWord(String toBePhonemised, boolean allowOtherLanguage) {
// cleanAllInstanceVariables();
Word currentWord = new Word(toBePhonemised);
// Vector result = new Vector();
String transcription = null;
Result currentResult = new Result();
long time1 = System.currentTimeMillis();
if (currentWord.getToBePhonemised().equals("")) {
logger.debug("Empty String!");
// return null;
return currentResult;
}
// word or item must at least have length 3 to be a denglish item
if (currentWord.getToBePhonemised().length() <= 2) {
logger.debug("Input to short to be a deng item");
logger.debug("Now using letter to sound rules");
// return null;
return currentResult;
}
// try cutting off inflection ending and/or prefix:
transcription = processFlection(currentWord, currentResult, allowOtherLanguage);
if (transcription != null) {
long time2 = System.currentTimeMillis();
long end = time2 - time1;
logger.debug("Processing took: " + end + " ms");
// return transcription;
currentResult.setTranscription(transcription);
// System.out.println("1) var is "+currentResult.isUsedOtherLanguageToPhonemise());
return currentResult;
}
// try compound analysis, first without, then with other language:
transcription = compoundAnalysis(currentWord, currentResult, false);
if (transcription != null) {
long time3 = System.currentTimeMillis();
long end = time3 - time1;
logger.debug("Processing took: " + end + " ms");
// return transcription;
currentResult.setTranscription(transcription);
// System.out.println("2) var is "+currentResult.isUsedOtherLanguageToPhonemise());
return currentResult;
}
transcription = compoundAnalysis(currentWord, currentResult, allowOtherLanguage);
if (transcription != null) {
long time3 = System.currentTimeMillis();
long end = time3 - time1;
logger.debug("Processing took: " + end + " ms");
// return transcription;
currentResult.setTranscription(transcription);
// System.out.println("2) var is "+currentResult.isUsedOtherLanguageToPhonemise());
return currentResult;
}
// return null;
return currentResult;
}
/**
* Try to process the input word, as it stands, or by cutting off prefixes or inflectional
* suffixes.
*
* @param toBePhonemised the input word
* @param allowOtherLanguage allowOtherLanguage
* @return the transcription of the word, or null if the word could not be transcribed
*/
private String processFlection(Word word, Result currentResult, boolean allowOtherLanguage) {
String toBePhonemised = word.getToBePhonemised();
logger.debug("processFlection is starting with: " + toBePhonemised);
// First of all, make sure there is no userdict/lexicon entry:
String transcription = jphon.userdictLookup(toBePhonemised, null);
if (transcription != null) {
return transcription;
}
transcription = jphon.lexiconLookup(toBePhonemised, null);
if (transcription != null) {
return transcription;
}
// Try to process by cutting off endings only, without cutting off prefix:
if (allowOtherLanguage) {
transcription = processFlectionEnding(word, currentResult);
}
if (transcription != null) {
return transcription;
}
// try removing prefix:
// Enforce at least 3 characters in the stem (the part of the word that comes after the prefix):
int maxPrefLen = Math.min(this.maxPrefixLength, word.getToBePhonemised().length() - 3);
for (int i = maxPrefLen; i > 0; i--) {
String prefix = word.getToBePhonemised().substring(0, i).toLowerCase();
String prefixPhon = prefixLexiconLookup(prefix);
if (prefixPhon != null) {
logger.debug("Prefix found: " + prefix + " [" + prefixPhon + "]");
Word partialWord = new Word(word.getToBePhonemised().substring(i));
// recursively call this method, i.e. allow multiple prefixes:
String restTranscription = processFlection(partialWord, currentResult, allowOtherLanguage);
if (restTranscription != null) { // yes, found valid analysis
if (prefixPhon.indexOf("'") != -1) {
restTranscription = restTranscription.replaceAll("'", "");
}
transcription = prefixPhon + "-" + restTranscription;
return transcription;
}
}
}
return null;
}
/**
* Try to process the input word as a verbal or adjective flection.
*
* @param word the input word
* @param currentResult currentResult
* @return the transcription of the word
*/
private String processFlectionEnding(Word word, Result currentResult) {
String toBePhonemised = word.getToBePhonemised();
logger.debug("processFlectionEnding is starting with: " + toBePhonemised);
String wordMinusFlectionEnding = null;
String flectionEnding = null;
String result = null;
// separateFlectionEndings returns null if no valid flection ending can be found
// otherwise it returns an array containing the word without flection and the flection
String[] wordPlusEnding = separateFlectionEndings(toBePhonemised, this.maxEndingLength);
if (wordPlusEnding != null) {
wordMinusFlectionEnding = wordPlusEnding[0];
flectionEnding = wordPlusEnding[1];
word =
transformWordToEnBaseForm(
wordMinusFlectionEnding, flectionEnding, word); // language-dependent
if (word.getOtherLanguageBaseForm() != null) {
word.setFlectionEnding(flectionEnding);
result = transcribeFlection(word, currentResult); // language-dependent
} else { // case of upgedatete
// start separateFlectionEndings() with a smaller number of ending chars
int currentEndingLength = flectionEnding.length();
wordPlusEnding = separateFlectionEndings(toBePhonemised, currentEndingLength - 1);
if (wordPlusEnding != null) {
wordMinusFlectionEnding = wordPlusEnding[0];
flectionEnding = wordPlusEnding[1];
word =
transformWordToEnBaseForm(
wordMinusFlectionEnding, flectionEnding, word); // language-dependent
if (word.getOtherLanguageBaseForm() != null) {
word.setFlectionEnding(flectionEnding);
result = transcribeFlection(word, currentResult); // language-dependent
} else {
currentEndingLength = flectionEnding.length();
wordPlusEnding = separateFlectionEndings(toBePhonemised, currentEndingLength - 1);
if (wordPlusEnding != null) {
wordMinusFlectionEnding = wordPlusEnding[0];
flectionEnding = wordPlusEnding[1];
word =
transformWordToEnBaseForm(
wordMinusFlectionEnding, flectionEnding, word); // language-dependent
if (word.getOtherLanguageBaseForm() != null) {
word.setFlectionEnding(flectionEnding);
result = transcribeFlection(word, currentResult); // language-dependent
}
}
}
} else { // array is null
// we have sth. that is already without flection ending like 'check'
word = transformWordToEnBaseForm(toBePhonemised, null, word); // language-dependent
if (word.getOtherLanguageBaseForm() != null) {
result = transcribeFlection(word, currentResult); // language-dependent
} else {
logger.debug("Unable to transcribe flection. Returning null.");
}
}
}
}
// System.out.println("var is in processFlection:
// "+currentResult.isUsedOtherLanguageToPhonemise());
logger.debug("processFlection: " + result);
return result;
}
/**
* Analyses parts of input word for affixes, compounds etc.
*
* @param word the input word
* @param currentResult currentResult
* @param allowOtherLanguage whether to allow component words from other language in compound
* analysis
* @return If a transcription for the input can be found, then it is returned. Otherwise returns
* null.
*/
private String compoundAnalysis(Word word, Result currentResult, boolean allowOtherLanguage) {
// Chop off longest possible prefixes and try to look them up
// in the lexicon. Any part must have a minimum length of 3 -> 2!! characters.
logger.debug("compoundAnalysis is starting with: " + word.getToBePhonemised());
for (int i = word.getToBePhonemised().length() - 3; i >= 3; i--) { // -3!!! >= 3!!!
String firstPhon = null;
String fugePhon = null;
String restPhon = null;
String[] genitiveAccusativeAndPluralEndings =
getEndingsAndAffixes("noun_genitive_accusative_and_plural_endings"); // should
// be
// 's'
// and
// 'n'
// for
// german
String prefix = word.getToBePhonemised().substring(0, i);
logger.debug("Pre: " + prefix);
firstPhon = jphon.userdictLookup(prefix, null);
if (firstPhon == null) {
firstPhon = jphon.lexiconLookup(prefix, null);
}
if (firstPhon == null && allowOtherLanguage) {
firstPhon = jphon.phonemiseEn(prefix);
if (firstPhon != null) {
currentResult.setUsedOtherLanguageToPhonemise(true);
}
}
if (firstPhon != null) { // found a valid prefix
// TODO: shouldn't this call processFlection()?
String rest = word.getToBePhonemised().substring(i);
logger.debug("Rest is: " + rest);
// Is the rest a simple lexical entry?
// restPhon = germanLexiconLookup(rest);
restPhon = prefixLexiconLookup(rest);
logger.debug("RestPhon: " + restPhon);
if (restPhon == null) {
restPhon = jphon.userdictLookup(rest, null);
}
if (restPhon == null) {
restPhon = jphon.lexiconLookup(rest, null);
}
if (restPhon == null && allowOtherLanguage) {
restPhon = jphon.phonemiseEn(rest);
if (restPhon != null) {
currentResult.setUsedOtherLanguageToPhonemise(true);
}
}
if (restPhon == null) {
for (int j = 0; j < genitiveAccusativeAndPluralEndings.length; j++) {
if (rest.endsWith(genitiveAccusativeAndPluralEndings[j])) {
logger.debug("rest ends with: " + genitiveAccusativeAndPluralEndings[j]);
String restWithoutLast = rest.substring(0, rest.length() - 1);
String restPhonDe = jphon.userdictLookup(restWithoutLast, null);
if (restPhonDe == null) restPhonDe = jphon.lexiconLookup(restWithoutLast, null);
String genitiveAndPluralEndingTrans =
endingTranscriptionLookup(genitiveAccusativeAndPluralEndings[j]);
if (restPhonDe != null) {
restPhon = restPhonDe + genitiveAndPluralEndingTrans;
} else if (allowOtherLanguage) {
String restPhonEn = jphon.phonemiseEn(rest.substring(0, rest.length() - 1));
if (restPhonEn != null) {
currentResult.setUsedOtherLanguageToPhonemise(true);
restPhon = restPhonEn + genitiveAndPluralEndingTrans;
}
}
}
if (restPhon != null) break;
}
}
// Or does it help if we cut off a Fuge?
if (restPhon == null) {
String[] helper = fugeSearch(rest);
if (helper != null && helper.length == 2) {
fugePhon = helper[0];
String rest2 = helper[1];
restPhon = jphon.userdictLookup(rest2, null);
if (restPhon == null) {
restPhon = jphon.lexiconLookup(rest2, null);
}
if (restPhon == null && allowOtherLanguage) {
restPhon = jphon.phonemiseEn(rest2);
if (restPhon != null) {
currentResult.setUsedOtherLanguageToPhonemise(true);
}
}
if (restPhon == null)
restPhon = compoundAnalysis(new Word(rest2), currentResult, allowOtherLanguage);
}
}
// Maybe rest is a flection
if (restPhon == null) {
// System.out.println("1) new word is : "+rest+". processFlection is called from here. var
// is : "+currentResult.isUsedOtherLanguageToPhonemise());
restPhon = processFlection(new Word(rest), currentResult, allowOtherLanguage);
// System.out.println("2) new word was : "+rest+". processFlection is called from here.
// var is : "+currentResult.isUsedOtherLanguageToPhonemise());
}
// Or can the rest be analysed as a compound?
if (restPhon == null)
restPhon = compoundAnalysis(new Word(rest), currentResult, allowOtherLanguage);
if (restPhon != null) {
// In restPhon, delete stress signs:
restPhon = restPhon.replaceAll("'", "");
return firstPhon + (fugePhon != null ? fugePhon : "") + "-" + restPhon;
}
}
}
return null;
}
/**
* Try to cut off a Fuge morpheme at the beginning of suffix.
*
* @param suffix a part of a word with a prefix already removed.
* @return a two-item String array. First string is the trannscnscription of the Fuge found,
* second is the suffix after the Fuge was removed. Returns null if no Fuge was found.
*/
private String[] fugeSearch(String suffix) {
String fugePhon = null;
int fugeLength = 0;
String[] validFuges = getEndingsAndAffixes("compound_fuge");
for (int j = 0; j < validFuges.length; j++) {
if (suffix.startsWith(validFuges[j])) {
fugePhon = endingTranscriptionLookup(validFuges[j]);
fugeLength = validFuges[j].length();
break;
}
}
if (fugePhon != null) { // found a Fuge
String[] returnValue = new String[2];
returnValue[0] = fugePhon;
returnValue[1] = suffix.substring(fugeLength);
return returnValue;
} else {
return null;
}
}
/**
* Separates flection ending from input word.
*
* @param toBePhonemised the input word
* @param endingLength endings from language specific ending list
* @return when valid flection ending is found, returns a string array of two elements (first is
* stem, second is ending); else, returns null.
*/
private String[] separateFlectionEndings(String toBePhonemised, int endingLength) {
String wordMinusFlectionEnding = null;
String flectionEnding = knowEnding(toBePhonemised, endingLength);
if (flectionEnding != null) {
String[] wordPlusEnding = new String[2];
wordMinusFlectionEnding =
toBePhonemised.substring(0, toBePhonemised.length() - flectionEnding.length());
wordPlusEnding[0] = wordMinusFlectionEnding;
wordPlusEnding[1] = flectionEnding;
return wordPlusEnding;
} else {
return null;
}
}
/**
* Try to find baseform of otherLanguageWord (i.e. english infinitive in denglish word)
*
* @param wordMinusFlectionEnding wordMinusFlectionEnding
* @param flectionEnding flectionEnding
* @param word word
* @return word
*/
private Word transformWordToEnBaseForm(
String wordMinusFlectionEnding, String flectionEnding, Word word) {
logger.debug("getEnBaseForm is starting with...: " + wordMinusFlectionEnding);
String[] participleBaseLong = getEndingsAndAffixes("participle_base_long"); // 'et' for german
String[] participleBaseShort = getEndingsAndAffixes("participle_base_short"); // 't' for german
// String[] flectionFuge = getEndingsAndAffixes("flection_fuge");//should be 'e' for german
String wordMinusFlectionEndingPenultimateChar = null;
String wordMinusFlectionEndingUltimateChar = null;
if (wordMinusFlectionEnding.length() > 2) {
wordMinusFlectionEndingPenultimateChar =
wordMinusFlectionEnding.substring(
wordMinusFlectionEnding.length() - 2, wordMinusFlectionEnding.length() - 1);
}
if (wordMinusFlectionEnding.length() > 1) {
wordMinusFlectionEndingUltimateChar =
wordMinusFlectionEnding.substring(
wordMinusFlectionEnding.length() - 1, wordMinusFlectionEnding.length());
}
String wordMinusFlectionEndingLastTwo =
wordMinusFlectionEndingPenultimateChar + wordMinusFlectionEndingUltimateChar;
if (wordMinusFlectionEnding.length() > 3) {
if (knowEnBaseForm(
wordMinusFlectionEnding)) { // item without ending is already en base form like >boot<
if (flectionEnding != null) {
if (isLongParticipleBaseEnding(flectionEnding, participleBaseLong)
|| isShortParticipleBaseEnding(
flectionEnding, participleBaseShort)) { // 'boot >et< or 'scroll>t<'
word.setOtherLanguageBaseForm(wordMinusFlectionEnding);
word.setCouldBeParticiple(true);
word.setCouldBeParticipleInBaseForm(true);
// next is special case for words like dat>e<>te<
} else if (endsWithVowel(wordMinusFlectionEnding)) { // 'te'
word.setOtherLanguageBaseForm(wordMinusFlectionEnding);
word.setWordMinusFlectionEndsWithVowel(true);
word.setCouldBeParticiple(true);
} else { // downloaden
word.setOtherLanguageBaseForm(wordMinusFlectionEnding);
}
} else { // scroll
word.setOtherLanguageBaseForm(wordMinusFlectionEnding);
logger.debug("wordMinusFlectionEnding is already enBaseForm");
}
}
// (up | ge) | date >t< (em) (j = 1)
// (ge) | boot >et< (em) (j =2)
// scroll >t< (en) (j=1)
// scan >nt< (en) (j=2)
if (word.getOtherLanguageBaseForm() == null) {
for (int j = 1; j < 3; j++) { // chop off 1-3 chars from end of word
logger.debug(
"new(2a): "
+ wordMinusFlectionEnding.substring(0, wordMinusFlectionEnding.length() - j));
if (knowEnBaseForm(
wordMinusFlectionEnding.substring(0, wordMinusFlectionEnding.length() - j))) {
if (isLongParticipleBaseEnding(
wordMinusFlectionEndingLastTwo, participleBaseLong) // 'et'
|| isShortParticipleBaseEnding(
wordMinusFlectionEndingUltimateChar, participleBaseShort)
&& !(isShortParticipleBaseEnding(
wordMinusFlectionEndingUltimateChar, participleBaseShort)
&& // to
// force
// that
// sth
// like
// 'cha>tt<en'
// is
// correctly
isShortParticipleBaseEnding(
wordMinusFlectionEndingPenultimateChar,
participleBaseShort))) { // processed
// in
// geminate
// clause
word.setOtherLanguageBaseForm(
wordMinusFlectionEnding.substring(0, wordMinusFlectionEnding.length() - j));
word.setCouldBeParticiple(true);
word.setCutOffCharacter(true);
logger.debug("new(2a)");
}
}
if (word.getOtherLanguageBaseForm() != null) break;
}
}
if (word.getOtherLanguageBaseForm() == null) {
// check for geminates -> scannen
logger.debug("in geminate clause: " + wordMinusFlectionEnding);
if (wordMinusFlectionEndingUltimateChar.equals(
wordMinusFlectionEndingPenultimateChar)) { // sca>nn<
if (knowEnBaseForm(
wordMinusFlectionEnding.substring(0, wordMinusFlectionEnding.length() - 1))) {
word.setOtherLanguageBaseForm(
wordMinusFlectionEnding.substring(0, wordMinusFlectionEnding.length() - 1));
logger.debug("geminate.......");
}
}
}
if (word.getOtherLanguageBaseForm() == null) {
// try to test if it is a gerund -> updatend
word.setIsVerbalGerund(checkIfGerund(wordMinusFlectionEnding));
if (word.getIsVerbalGerund()) { // we have a gerund
word.setOtherLanguageBaseForm(transformWordToEnBaseFormGerund(wordMinusFlectionEnding));
}
}
}
logger.debug("finally enBaseForm: " + word.getOtherLanguageBaseForm());
return word;
}
/**
* @param word word
* @return enBaseForm
*/
private String transformWordToEnBaseFormGerund(String word) {
logger.debug("getBaseFormGerund called with: " + word);
String enBaseForm = null;
String[] flectionFuge = getEndingsAndAffixes("flection_fuge"); //
logger.debug("found gerund..........");
for (int j = 0; j < flectionFuge.length; j++) {
if (knowEnBaseForm(word.substring(0, word.length() - 3) + flectionFuge[j])) { // updat>e<nd
enBaseForm = word.substring(0, word.length() - 3) + flectionFuge[j]; // like updat >e< nd
logger.debug("gerund case 3");
}
if (enBaseForm != null) break;
}
if (enBaseForm == null) {
if (knowEnBaseForm(word.substring(0, word.length() - 3))) { // download>end<
enBaseForm = word.substring(0, word.length() - 3); // item without 'end' is base
logger.debug("gerund case 1");
} else if (knowEnBaseForm(word.substring(0, word.length() - 4))
&& word.charAt(word.length() - 4) == word.charAt(word.length() - 5)) { // scan>n< end
enBaseForm = word.substring(0, word.length() - 4);
logger.debug("gerund case 2");
}
}
return enBaseForm;
}
/**
* Building the transcription and syllabification of a flection
*
* @param currentResult currentResult
* @param word : the English infinitive as found in English lexicon
* @return transcription of complete input word
*/
private String transcribeFlection(Word word, Result currentResult) {
String result = null;
String otherLanguageTranscription = null;
String endingTranscription = null;
String gerundEndingTrans = null;
String participleBaseShortEndingTrans = null;
String flectionFugeTrans = null;
otherLanguageTranscription = jphon.phonemiseEn(word.getOtherLanguageBaseForm());
if (otherLanguageTranscription != null) {
// System.out.println("var should be true");
currentResult.setUsedOtherLanguageToPhonemise(true);
for (int j = 0; j < this.dentalPlosives.length; j++) {
if (otherLanguageTranscription.endsWith(this.dentalPlosives[j])) {
word.setExtraSyll(true);
logger.debug("extraSyll true");
}
}
// System.out.println("var is in transcribeFlection:
// "+currentResult.isUsedOtherLanguageToPhonemise());
// for cases like 'scrollet' where 'et' is flection ending and NOT ending of
// participleBaseForm; otherwise 'scrollet' would sound like 'scrollt'
String[] participleBaseLongEndings = getEndingsAndAffixes("participle_base_long");
for (int j = 0; j < participleBaseLongEndings.length; j++) {
if (word.getFlectionEnding() != null
&& word.getFlectionEnding().equals(participleBaseLongEndings[j])
&& !(word.getCutOffCharacter())) { // 'et'
word.setExtraSyll(true);
}
}
String[] gerundEndings = getEndingsAndAffixes("gerund_ending"); // should be 'end' -> bootend
// String gerundEndingTrans = endingTranscriptionLookup(gerundEnding);//should be '@nt'
for (int j = 0; j < gerundEndings.length; j++) {
if (endingTranscriptionLookup(gerundEndings[j]) != null) {
gerundEndingTrans = endingTranscriptionLookup(gerundEndings[j]);
}
}
String[] participleBaseShortEndings = getEndingsAndAffixes("participle_base_short");
// If the participle ends with 'ed' or 'et' doesn't matter -> you get the same transcription
// String participleBaseEndingTrans = endingTranscriptionLookup(participleBaseEnding);//gives
// you 't'
for (int j = 0; j < participleBaseShortEndings.length; j++) {
if (endingTranscriptionLookup(participleBaseShortEndings[j]) != null) {
participleBaseShortEndingTrans =
endingTranscriptionLookup(participleBaseShortEndings[j]); // gives you 't'
}
}
String[] flectionFuge = getEndingsAndAffixes("flection_fuge"); // gives you 'e'
for (int j = 0; j < flectionFuge.length; j++) {
if (endingTranscriptionLookup(flectionFuge[j]) != null) {
flectionFugeTrans = endingTranscriptionLookup(flectionFuge[j]);
}
}
endingTranscription = endingTranscriptionLookup(word.getFlectionEnding());
String newEnTranscription = rebuildTrans(otherLanguageTranscription);
String newGerundEndingTrans = rebuildTrans(gerundEndingTrans); // should then be '@n-t'
String voicedNewGerundEndingTrans = voiceFinal(newGerundEndingTrans); // should be '@n-d'
// String voicedGerundEndingTrans = voiceFinal(gerundEndingTrans); //should be '@nd'
logger.debug("enTrans: " + otherLanguageTranscription);
if (word.getFlectionEnding() != null) {
if (endingTranscriptionLookup(word.getFlectionEnding()) != null) {
// special rule in case of enBaseForm's last char equals valid flection ending i.e. 't'
// in this case give us back the enBaseForm aka enInfinitive
// testing for participle because of date>te< enBaseForm ends with found ending
if (otherLanguageTranscription.endsWith(word.getFlectionEnding())
&& !(word.getIsVerbalGerund())
&& !(word.getCouldBeParticiple())) {
result = otherLanguageTranscription;
logger.debug("(0)");
} else {
if (word.getCouldBeParticiple()
&& isShortSuperlative(word.getFlectionEnding())
&& word.getExtraSyll()) { // i.e.
// downgeloadetsten
result =
newEnTranscription
+ flectionFugeTrans
+ participleBaseShortEndingTrans
+ endingTranscription;
logger.debug("(1)");
} else if (word.getCouldBeParticiple()
&& word.getCouldBeParticipleInBaseForm()
&& word.getExtraSyll()) { // scrollet
// or
// downloadet
result = newEnTranscription + flectionFugeTrans + participleBaseShortEndingTrans;
logger.debug("(2)");
} else if (word.getCouldBeParticiple()
&& word.getExtraSyll()
&& word.getWordMinusFlectionEndsWithVowel()) {
result = newEnTranscription + flectionFugeTrans + "-" + endingTranscription;
logger.debug("(3)");
} else if (word.getCouldBeParticiple() && word.getExtraSyll()) { // i.e. downgeloadetere
result =
newEnTranscription
+ flectionFugeTrans
+ "-"
+ participleBaseShortEndingTrans
+ endingTranscription;
logger.debug("(4)");
} else if (word.getCouldBeParticiple()
&& isShortSuperlative(word.getFlectionEnding())) { // i.e.
// gescrolltstem
result =
otherLanguageTranscription + participleBaseShortEndingTrans + endingTranscription;
logger.debug("(5)");
} else if (word.getCouldBeParticiple() && word.getCouldBeParticipleInBaseForm()) {
result = otherLanguageTranscription + participleBaseShortEndingTrans;
logger.debug("(6)");
} else if (word.getCouldBeParticiple()) { // i.e. gescrolltestem
result =
otherLanguageTranscription
+ "-"
+ participleBaseShortEndingTrans
+ endingTranscription;
logger.debug("(7)");
} else {
if (word.getIsVerbalGerund()) {
logger.debug("isVerbalGerund");
if (isShortSuperlative(word.getFlectionEnding())) {
result = newEnTranscription + gerundEndingTrans + endingTranscription;
} else {
result = newEnTranscription + voicedNewGerundEndingTrans + endingTranscription;
}
} else {
if (isShortSuperlative(word.getFlectionEnding())) {
result = otherLanguageTranscription + endingTranscription;
} else { // no Gerund, no superlative but maybe something like 'scannst'
if (word.getExtraSyll()) { // means: word ends on 't' or 'd'
logger.debug("extraSyll is true here...");
result = newEnTranscription + endingTranscription;
} else { // means: word ends on something else
if (endingContainsVowel(word.getFlectionEnding())
&& (!(endingBeginsWithVowel(word.getFlectionEnding())))) {
result = otherLanguageTranscription + "-" + endingTranscription;
} else {
if (endingContainsVowel(word.getFlectionEnding())
&& endingBeginsWithVowel(word.getFlectionEnding())) {
result = newEnTranscription + endingTranscription;
} else {
result = otherLanguageTranscription + endingTranscription;
}
}
}
}
}
}
}
}
} else { // flection ending is null: two possibilities: en-Word like boot or ger gerund like
// bootend
if (word.getIsVerbalGerund()) {
result = newEnTranscription + gerundEndingTrans;
logger.debug("(((1)))");
} else { // scann, date
result = otherLanguageTranscription;
logger.debug("(((2)))");
}
}
}
return result;
}
/**
* Checks if input word has valid ending from ending list.
*
* @param toBePhonemised The input word
* @param endingLength endingLength
* @return Flection ending if one can be found, null otherwise
*/
private String knowEnding(String toBePhonemised, int endingLength) {
logger.debug("in knowEnding: " + toBePhonemised);
String currentEnding = null;
String foundEnding = null;
int wordLength = toBePhonemised.length();
for (int j = endingLength; j > 0; j--) {
if (j < wordLength) {
currentEnding = toBePhonemised.substring(wordLength - j, wordLength);
logger.debug("currentEnding: " + currentEnding);
if (this.endingSet.contains(currentEnding)) {
foundEnding = currentEnding;
logger.debug("foundEnding....: " + foundEnding);
}
} else {
continue;
}
if (foundEnding != null) break;
}
return foundEnding;
}
/**
* Voices the final consonant of an ending.
*
* @param ending the input ending
* @return voiced ending of ending can be voiced, input ending otherwise
*/
private String voiceFinal(String ending) {
String finalPhoneme = null;
String voicedFinalPhoneme = null;
String voicedEnding = null;
if (ending.length() > 0) {
finalPhoneme = ending.substring(ending.length() - 1, ending.length());
}
if (getVoicedFinal(finalPhoneme) != null) {
voicedFinalPhoneme = getVoicedFinal(finalPhoneme);
voicedEnding =
ending.substring(0, ending.length() - 1) + voicedFinalPhoneme; // return new ending voiced
} else {
// if there is no voiced value for the last phone, return ending as it came in
voicedEnding = ending;
}
return voicedEnding;
}
/**
* If the given string ends with a consonant, insert a syllable boundary before that consonant.
* Otherwise, append a syllable boundary.
*
* @param s input syllable
* @return syllable with boundaries reset
*/
private String rebuildTrans(String s) {
AllophoneSet set = jphon.getAllophoneSet();
if (set != null) {
Allophone[] allophones = set.splitIntoAllophones(s);
if (allophones != null && allophones.length > 0) {
Allophone last = allophones[allophones.length - 1];
if (last.isConsonant()) { // insert a syllable boundary before final consonant
String lastPh = last.name();
return s.substring(0, s.length() - lastPh.length()) + "-" + lastPh;
}
}
}
return s + "-";
}
/**
* Checks if input string is a gerund - means: input ends with 'end' Sets global var
* isVerbalGerund which is important for building transcription later on
*
* @param s input string
* @return true if s.substring(s.length -3, s.length).equals(gerundEndings[j]), false otherwise
*/
private boolean checkIfGerund(String s) {
String[] gerundEndings = getEndingsAndAffixes("gerund_ending"); // should be 'end' for german
for (int j = 0; j < gerundEndings.length; j++) {
if (s.length() > gerundEndings[j].length()) {
if (s.substring(s.length() - 3, s.length()).equals(gerundEndings[j])) { // we have an gerund
return true;
}
}
}
return false;
}
/**
* Checks if flection ending is a short superlative ending
*
* @param flectionEnding flection ending
* @return true if ending is short superlative false otherwise
*/
private boolean isShortSuperlative(String flectionEnding) {
String[] shortSuperlativeEndings = getEndingsAndAffixes("superlative_short");
for (int i = 0; i < shortSuperlativeEndings.length; i++) {
if (flectionEnding != null) {
if (flectionEnding.equals(shortSuperlativeEndings[i])) {
return true;
}
}
}
return false;
}
/**
* Checks if flection ending begins with a vowel
*
* @param ending flection ending
* @return true if ending begins with a vowel, false otherwise
*/
private boolean endingBeginsWithVowel(String ending) {
for (int i = 0; i < this.vowels.length; i++) {
if (this.vowels[i].equals(ending.substring(0, 1))) {
return true;
}
}
return false;
}
/**
* Checks if flection ending contains a vowel
*
* @param ending flection ending
* @return true if ending contains a vowel, false otherwise
*/
private boolean endingContainsVowel(String ending) {
for (int i = 0; i < this.vowels.length; i++) {
for (int j = 0; j < ending.length(); j++) {
if (this.vowels[i].equals(ending.substring(j, j + 1))) {
return true;
}
}
}
return false;
}
private boolean endsWithVowel(String s) {
for (int i = 0; i < this.vowels.length; i++) {
for (int j = 0; j < s.length(); j++) {
if (this.vowels[i].equals(s.substring(s.length() - 1, s.length()))) {
return true;
}
}
}
return false;
}
private boolean isLongParticipleBaseEnding(String s, String[] participleBaseLong) {
for (int j = 0; j < participleBaseLong.length; j++) {
if (s.equals(participleBaseLong[j])) { // 'et'
return true;
}
}
return false;
}
private boolean isShortParticipleBaseEnding(String s, String[] participleBaseShort) {
for (int j = 0; j < participleBaseShort.length; j++) {
if (s.equals(participleBaseShort[j])) { // 't'
return true;
}
}
return false;
}
public Locale getLocale() {
return this.locale;
}
/**
* Checks if item is in english lexicon.
*
* @param s english base form
* @return true if item is in english lexicon, false if not
*/
private boolean knowEnBaseForm(String s) {
if (jphon.phonemiseEn(s) != null) {
return true;
}
return false;
}
// public boolean usedOtherLanguageToPhonemise(boolean usedOtherLanguageToPhonemise) {
// return usedOtherLanguageToPhonemise;
// }
/**
* ***************************************************************************** Hashtable lookup
* methods for morphology & phonology information
* *****************************************************************************
*/
/**
* Looking for item in german prefix lexicon
*
* @param s item to be found
* @return Transcription of item
*/
private String prefixLexiconLookup(String s) {
String prefixTranscription = null;
try {
if (this.prefixLexicon.get(s) != null) {
prefixTranscription = (String) this.prefixLexicon.get(s);
}
} catch (Exception e) {
logger.debug("prefixLexLookup: " + e.toString());
}
return prefixTranscription;
}
/**
* Looks up in terminal voicing hash if there is a match for a unvoiced consonant
*
* @param phon The unvoiced consonant
* @return Voiced consonant if one can be found, null otherwise
*/
private String getVoicedFinal(String phon) {
String voicedPhon = null;
try {
if (this.terminalVoicings.get(phon) != null) {
voicedPhon = (String) this.terminalVoicings.get(phon);
}
} catch (Exception e) {
logger.debug("getVoicedFinal: " + e.toString());
}
return voicedPhon;
}
/**
* Looks up list of all endings and affixes for specific language
*
* @param key key of the affixes/endings you want to get, i.e. superlative_short
* @return list of endings or affixes if key is valid, null otherwise
*/
private String[] getEndingsAndAffixes(String key) {
String[] endingList = null;
try {
if (this.endingsAndAffixes.get(key) != null) {
String value = (String) this.endingsAndAffixes.get(key);
endingList = value.split("/");
}
} catch (Exception e) {
logger.debug("getEndingsAndAffixes: " + e.toString());
}
return endingList;
}
/**
* Try to get transcription for ending
*
* @param s The ending to be phonemised
* @return Transcription of ending
*/
private String endingTranscriptionLookup(String s) {
String affixPhon = null;
try {
if (this.flectionToPhon.get(s) != null) {
affixPhon = (String) this.flectionToPhon.get(s);
}
} catch (Exception e) {
logger.debug("endingTranscriptionLookup: " + e.toString());
}
return affixPhon;
}
}
/**
* A representation of any type of mary data, be it input, intermediate or output data. The
* "technical" representation of the read data is hidden from the caller, but can be accessed on
* request. Internally, the data is appropriately represented according to this data's type, i.e. as
* a String containing plain text, an XML DOM tree, or an input stream containing audio data.
*
* @author Marc Schröder
*/
public class MaryData {
private MaryDataType type;
private Locale locale;
private String outputParams = null;
// Only one of the following data representations should be non-null
// for a given instance; which one depends on our type.
private Document xmlDocument = null;
private String plainText = null;
private AudioInputStream audio = null;
private AudioFileFormat audioFileFormat = null;
private Logger logger = MaryUtils.getLogger("IO");
// for plainText, allow additional information:
private Voice defaultVoice = null;
private String defaultStyle = "";
private String defaultEffects = "";
// The following XML I/O helpers are only initialised
// if actually needed.
private MaryNormalisedWriter writer = null;
private boolean doValidate;
private boolean doWarnClient = false;
public MaryData(MaryDataType type, Locale locale) {
this(type, locale, false);
}
public MaryData(MaryDataType type, Locale locale, boolean createStubDocument) {
if (type == null) throw new NullPointerException("Received null type for MaryData");
this.type = type;
this.locale = locale;
// The following is the default setting for module output (we suppose
// that for the input data, setValidating() is called as appropriate):
doValidate = MaryProperties.getBoolean("maryxml.validate.modules", false);
if (createStubDocument && type.isMaryXML()) {
xmlDocument = MaryXML.newDocument();
}
}
public boolean getValidating() {
return doValidate;
}
public void setValidating(boolean doValidate) throws ParserConfigurationException {
this.doValidate = doValidate;
}
@Deprecated
public boolean getWarnClient() {
return doWarnClient;
}
@Deprecated
public void setWarnClient(boolean doWarnClient) {}
public MaryDataType getType() {
return type;
}
public Locale getLocale() {
return locale;
}
/**
* Read data from input stream <code>is</code>, in the appropriate way as determined by our <code>
* type</code>.
*
* @param is is
* @throws ParserConfigurationException ParserConfigurationException
* @throws SAXException SAXException
* @throws IOException IOException
* @throws TransformerConfigurationException TransformerConfigurationException
* @throws TransformerException TransformerException
*/
public void readFrom(InputStream is)
throws ParserConfigurationException, SAXException, IOException,
TransformerConfigurationException, TransformerException {
readFrom(is, null);
}
/**
* Read data from input stream <code>is</code>, in the appropriate way as determined by our <code>
* type</code>.
*
* @param is the InputStream from which to read.
* @param endMarker a string marking end of file. If this is null, read until end-of-file; if it
* is non-null, read up to (and including) the first line containing the end marker string.
* This will be ignored for audio data.
* @throws ParserConfigurationException ParserConfigurationException
* @throws SAXException SAXException
* @throws IOException IOException
* @throws TransformerConfigurationException TransformerConfigurationException
* @throws TransformerException TransformerException
*/
public void readFrom(InputStream is, String endMarker)
throws ParserConfigurationException, SAXException, IOException,
TransformerConfigurationException, TransformerException {
if (type.isXMLType() || type.isTextType())
readFrom(new InputStreamReader(is, "UTF-8"), endMarker);
else { // audio
// ignore endMarker
setAudio((AudioInputStream) is);
}
}
/**
* Read data from reader <code>r</code> in the appropriate way as determined by our <code>type
* </code>. Only XML and Text data can be read from a reader, audio data cannot.
*
* @param from from
* @throws ParserConfigurationException ParserConfigurationException
* @throws SAXException SAXException
* @throws IOException IOException
*/
public void readFrom(Reader from) throws ParserConfigurationException, SAXException, IOException {
String inputData = FileUtils.getReaderAsString(from);
setData(inputData);
}
/**
* Read data from reader <code>r</code> in the appropriate way as determined by our <code>type
* </code>. Only XML and Text data can be read from a reader, audio data cannot. "Helpers" needed
* to read the data, such as XML parser objects, are created when they are needed. If doWarnClient
* is set to true, warning and error messages related to XML parsing are logged to the log
* category connected to the client from which this request originated.
*
* @param from the Reader from which to read.
* @param endMarker a string marking end of file. If this is null, read until end-of-file; if it
* is non-null, read up to (and including) the first line containing the end marker string.
* @throws ParserConfigurationException ParserConfigurationException
* @throws SAXException SAXException
* @throws IOException IOException
*/
public void readFrom(Reader from, String endMarker)
throws ParserConfigurationException, SAXException, IOException {
// For the case that the data to be read it is not
// followed by end-of-file, we use a ReaderSplitter which
// provides a reader artificially "inserting" an end-of-file
// after a line containing the pattern given in endMarker.
Reader r = from;
if (endMarker != null) {
ReaderSplitter fromSplitter = new ReaderSplitter(from, endMarker);
r = fromSplitter.nextReader();
}
readFrom(r);
}
/**
* Set the content data of this MaryData object from the given String. For XML data ({@link
* MaryDataType#isXMLType()}), parse the String representation of the data into a DOM tree.
*
* @param dataString string representation of the input data.
* @throws ParserConfigurationException ParserConfigurationException
* @throws IOException IOException
* @throws SAXException SAXException
* @throws IllegalArgumentException if this method is called for MaryDataTypes that are neither
* text nor XML.
*/
public void setData(String dataString)
throws ParserConfigurationException, SAXException, IOException {
// First, some data cleanup:
dataString = StringUtils.purgeNonBreakingSpaces(dataString);
// Now, deal with it.
if (type.isXMLType()) {
logger.debug(
"Parsing XML input (" + (doValidate ? "" : "non-") + "validating): " + dataString);
xmlDocument = DomUtils.parseDocument(dataString, doValidate);
} else if (type.isTextType()) {
logger.debug("Setting text input: " + dataString);
plainText = dataString;
} else {
throw new IllegalArgumentException("Cannot set data of type " + type + " from a string");
}
}
/**
* Write our internal representation to output stream <code>os</code>, in the appropriate way as
* determined by our <code>type</code>.
*
* @param os os
* @throws TransformerConfigurationException TransformerConfigurationException
* @throws FileNotFoundException FileNotFoundException
* @throws TransformerException TransformerException
* @throws IOException IOException
* @throws Exception Exception
*/
public void writeTo(OutputStream os)
throws TransformerConfigurationException, FileNotFoundException, TransformerException,
IOException, Exception {
if (type.isXMLType()) {
if (writer == null) writer = new MaryNormalisedWriter();
if (logger.getEffectiveLevel().equals(Level.DEBUG)) {
ByteArrayOutputStream debugOut = new ByteArrayOutputStream();
writer.output(xmlDocument, debugOut);
logger.debug(debugOut.toString());
}
writer.output(xmlDocument, new BufferedOutputStream(os));
} else if (type.isTextType()) { // caution: XML types are text types!
writeTo(new OutputStreamWriter(os, "UTF-8"));
} else { // audio
logger.debug("Writing audio output, frame length " + audio.getFrameLength());
AudioSystem.write(audio, audioFileFormat.getType(), os);
os.flush();
os.close();
}
}
/*
* public void writeTo(HttpResponse response) throws TransformerConfigurationException, FileNotFoundException,
* TransformerException, IOException, Exception { if (type.isUtterances()) throw new
* IOException("Cannot write out utterance-based data type!");
*
* if (type.isXMLType()) { if (writer == null) writer = new MaryNormalisedWriter(); if
* (logger.getEffectiveLevel().equals(Level.DEBUG)) { ByteArrayOutputStream debugOut = new ByteArrayOutputStream();
* writer.output(xmlDocument, debugOut); logger.debug(debugOut.toString()); }
*
* //writer.output(xmlDocument, new BufferedOutputStream(os));
*
* ByteArrayOutputStream os = new ByteArrayOutputStream(); writer.output(xmlDocument, new BufferedOutputStream(os));
* NByteArrayEntity body = new NByteArrayEntity(os.toByteArray()); body.setContentType("text/html; charset=UTF-8");
* response.setEntity(body); } else if (type.isTextType()) // caution: XML types are text types! { //writeTo(new
* OutputStreamWriter(os, "UTF-8"));
*
* ByteArrayOutputStream os = new ByteArrayOutputStream(); writeTo(new OutputStreamWriter(os, "UTF-8")); NByteArrayEntity body
* = new NByteArrayEntity(os.toByteArray()); body.setContentType("text/html; charset=UTF-8"); response.setEntity(body); } else
* // audio { logger.debug("Writing audio output, frame length "+audio.getFrameLength()); //AudioSystem.write(audio,
* audioFileFormat.getType(), os); //os.flush();
*
* ByteArrayOutputStream os = new ByteArrayOutputStream(); AudioSystem.write(audio, audioFileFormat.getType(), os);
* os.flush();
*
* MaryHttpServerUtils.toHttpResponse(os.toByteArray(), response); } }
*/
/**
* Write our internal representation to writer <code>w</code>, in the appropriate way as
* determined by our <code>type</code>. Only XML and Text data can be written to a writer, audio
* data cannot. "Helpers" needed to read the data, such as XML parser objects, are created when
* they are needed.
*
* @param w w
* @throws TransformerConfigurationException TransformerConfigurationException
* @throws FileNotFoundException FileNotFoundException
* @throws TransformerException TransformerException
* @throws IOException IOException
* @throws Exception Exception
*/
public void writeTo(Writer w)
throws TransformerConfigurationException, FileNotFoundException, TransformerException,
IOException, Exception {
if (type.isXMLType()) {
throw new IOException("Better write XML data to an OutputStream, not to a Writer");
} else if (type.isTextType()) { // caution: XML types are text types!
w.write(plainText);
w.flush();
logger.debug("Writing Text output:\n" + plainText);
} else { // audio - cannot write this to a writer
throw new Exception("Illegal attempt to write audio data to a character Writer");
}
}
public Object getData() {
if (type.isXMLType()) {
return xmlDocument;
} else if (type.isTextType()) {
return plainText;
} else { // audio
return audio;
}
}
public String getPlainText() {
return plainText;
}
public void setPlainText(String plainText) {
this.plainText = plainText;
}
public Document getDocument() {
return xmlDocument;
}
public void setDocument(Document xmlDocument) {
this.xmlDocument = xmlDocument;
}
public AudioInputStream getAudio() {
return audio;
}
/**
* Set the audio data. This will discard any previously set audio data. If audio data is to be
* appended, consider appendAudio().
*
* @param audio audio
*/
public void setAudio(AudioInputStream audio) {
this.audio = audio;
}
public void setDefaultVoice(Voice voice) {
if (voice == null) {
return;
}
// check that voice locale fits before accepting the voice:
Locale voiceLocale = null;
voiceLocale = voice.getLocale();
Locale docLocale = getLocale();
if (docLocale == null
&& getType().isXMLType()
&& getDocument() != null
&& getDocument().getDocumentElement().hasAttribute("xml:lang")) {
docLocale =
MaryUtils.string2locale(getDocument().getDocumentElement().getAttribute("xml:lang"));
}
if (docLocale != null
&& voiceLocale != null
&& !(MaryUtils.subsumes(docLocale, voiceLocale)
|| MaryUtils.subsumes(voiceLocale, docLocale))) {
logger.warn(
"Voice `"
+ voice.getName()
+ "' does not match document locale `"
+ docLocale
+ "' -- ignoring!");
}
this.defaultVoice = voice;
}
public Voice getDefaultVoice() {
return defaultVoice;
}
public void setDefaultStyle(String style) {
defaultStyle = style;
}
public String getDefaultStyle() {
return defaultStyle;
}
public void setDefaultEffects(String effects) {
defaultEffects = effects;
}
public String getDefaultEffects() {
return defaultEffects;
}
/**
* The audio file format is required only for data types serving as input to modules producing
* AUDIO data (e.g., MBROLA data), as well as for the AUDIO data itself. It should be set by the
* calling code before passing the data to the module producing AUDIO data.
*
* @param audioFileFormat audioFileFormat
*/
public void setAudioFileFormat(AudioFileFormat audioFileFormat) {
this.audioFileFormat = audioFileFormat;
}
public AudioFileFormat getAudioFileFormat() {
return audioFileFormat;
}
public void append(MaryData md) {
if (md == null) throw new NullPointerException("Received null marydata");
if (!md.getType().equals(this.getType()))
throw new IllegalArgumentException(
"Cannot append mary data of type `"
+ md.getType().name()
+ "' to mary data of type `"
+ this.getType().name()
+ "'");
if (getType().isXMLType()) {
NodeList kids = md.getDocument().getDocumentElement().getChildNodes();
logger.debug("Appending " + kids.getLength() + " nodes to MaryXML structure");
Element docEl = this.getDocument().getDocumentElement();
for (int i = 0; i < kids.getLength(); i++) {
docEl.appendChild(this.getDocument().importNode(kids.item(i), true));
}
} else if (getType().isTextType()) {
// Attention: XML type is a text type!
if (this.plainText == null) {
this.plainText = md.getPlainText();
} else {
this.plainText = this.plainText + "\n\n" + md.getPlainText();
}
} else if (getType().equals(MaryDataType.get("AUDIO"))) {
appendAudio(md.getAudio());
} else {
throw new UnsupportedOperationException(
"Cannot append two mary data items of type `" + getType() + "'");
}
}
/**
* For audio data, append more audio data to the one currently present. If no audio data is set
* yet, this call is equivalent to setAudio().
*
* @param audioToAppend the new audio data to append
*/
public void appendAudio(AudioInputStream audioToAppend) {
if (this.audio == null) setAudio(audioToAppend);
else if (this.audio instanceof AppendableSequenceAudioInputStream)
((AppendableSequenceAudioInputStream) this.audio).append(audioToAppend);
else
this.audio =
new SequenceAudioInputStream(
this.audio.getFormat(),
Arrays.asList(new AudioInputStream[] {this.audio, audioToAppend}));
}
public void setOutputParams(String params) {
this.outputParams = params;
}
public String getOutputParams() {
return outputParams;
}
public String toString() {
return Objects.toStringHelper(this)
.add("type", getType())
.add("locale", getLocale())
.add("output parameters", getOutputParams())
.add("data", getData())
.add("document", DomUtils.serializeToString(getDocument()))
.add("validating", getValidating())
.add("plain text", getPlainText())
.add("audio", getAudio())
.add("audio file format", getAudioFileFormat())
.toString();
}
}
/**
* An expansion pattern implementation for date patterns.
*
* <p>Modified from the German version of Marc Schröder with - removed all word abbreviations,
* month names are always numerical - moved up the YYYY MM DD pattern as it's most common - added
* expansion for day name
*/
public class DateEP extends ExpansionPattern {
private final String[] _knownTypes = {
"date",
"date:dmy",
"date:ymd", // standard Korean way is YY MM DD
"date:dm",
"date:my",
"date:y",
"date:m",
"date:d",
"date:mdy",
"date:md"
};
/**
* Every subclass has its own list knownTypes, an internal string representation of known types.
* These are possible values of the <code>type</code> attribute to the <code>say-as</code>
* element, as defined in MaryXML.dtd. If there is more than one known type, the first type (
* <code>knownTypes[0]</code>) is expected to be the most general one, of which the others are
* specialisations.
*/
private final List<String> knownTypes = Arrays.asList(_knownTypes);
public List<String> knownTypes() {
return knownTypes;
}
// Domain-specific primitives:
protected final String sDay = "(?:0?[1-9]|[12][0-9]|3[01])";
protected final String sMonth = "(?:0[1-9]|1[0-2])";
protected final String sYear = "(?:([0-9][0-9])?[0-9][0-9])";
protected final String sAbbrevDayPattern = "(?:[윌화수목금토일])";
protected final String sDot = "[\\.]";
protected final String sSlash = "[/]";
protected final String sMinus = "[\\-]";
protected final String sMatchingChars = "[0-9/\\.\\w\\(\\)윌화수목금토일]";
protected final String sSeparators = (sDot + "|" + sSlash + "|" + sMinus);
protected static Map<String, String> dayOfWeekExpansion = new HashMap<String, String>();
static {
dayOfWeekExpansion.put("윌", "윌요일");
dayOfWeekExpansion.put("화", "화요일");
dayOfWeekExpansion.put("수", "수요일");
dayOfWeekExpansion.put("목", "목요일");
dayOfWeekExpansion.put("금", "금요일");
dayOfWeekExpansion.put("토", "토요일");
dayOfWeekExpansion.put("일", "일요일");
}
// Now the actual match patterns:
protected final Pattern reDOWYearMonthDay =
Pattern.compile(
"(?:\\s|\\()*"
+ "("
+ sAbbrevDayPattern
+ ")"
+ "(?:"
+ sDot
+ "|"
+ sSlash
+ "|\\s|\\))*"
+ "("
+ sYear
+ ")"
+ "(?:"
+ sDot
+ "|"
+ sSlash
+ ")"
+ "("
+ sMonth
+ ")"
+ "(?:"
+ sDot
+ "|"
+ sSlash
+ ")"
+ "("
+ sDay
+ ")");
protected final Pattern reYearMonthDayDOW =
Pattern.compile(
"("
+ sYear
+ ")"
+ "(?:"
+ sDot
+ "|"
+ sSlash
+ ")"
+ "("
+ sMonth
+ ")"
+ "(?:"
+ sDot
+ "|"
+ sSlash
+ ")"
+ "("
+ sDay
+ ")"
+ "(?:"
+ sDot
+ "|"
+ sSlash
+ "|\\s|\\()*"
+ "("
+ sAbbrevDayPattern
+ ")"
+ "(?:\\s|\\))*"); // the list of allowed characters should prevent already expanded
// days of week tokens from being expanded again
protected final Pattern reYearMonthDay =
Pattern.compile(
"(" + sYear + ")" + "(?:" + sDot + "|" + sSlash + ")" + "(" + sMonth + ")" + "(?:" + sDot
+ "|" + sSlash + ")" + "(" + sDay + ")");
protected final Pattern reDayMonthYear =
Pattern.compile(
"(" + sDay + ")" + "(?:" + sDot + "|" + sSlash + ")" + "(" + sMonth + ")" + "(?:" + sDot
+ "|" + sSlash + ")" + "(" + sYear + ")");
protected final Pattern reMonthDayYear =
Pattern.compile(
"("
+ sMonth
+ ")"
+ "(?:"
+ sSeparators
+ ")"
+ "("
+ sDay
+ ")"
+ "(?:"
+ sSeparators
+ ")"
+ "("
+ sYear
+ ")");
protected final Pattern reDayMonth =
Pattern.compile(
"(" + sDay + ")" + "(?:" + sDot + "|" + sSlash + ")" + "(" + sMonth + ")" + "(?:" + sDot
+ "|" + sSlash + ")");
protected final Pattern reMonthDay =
Pattern.compile(
"("
+ sMonth
+ ")"
+ "(?:"
+ sSeparators
+ ")"
+ "("
+ sDay
+ ")"
+ "(?:"
+ sSeparators
+ ")");
protected final Pattern reMonthYear =
Pattern.compile("(" + sMonth + ")(" + sSlash + "|" + sDot + ")(" + sYear + ")");
protected final Pattern reYear = Pattern.compile("(?:([0-9][0-9])?[0-9][0-9])");
protected final Pattern reMonth = Pattern.compile("(" + sMonth + ")" + "(?:" + sDot + ")");
protected final Pattern reDay = Pattern.compile("(" + sDay + ")" + "?:" + sDot);
private final Pattern reMatchingChars = Pattern.compile(sMatchingChars);
public Pattern reMatchingChars() {
return reMatchingChars;
}
/**
* Every subclass has its own logger. The important point is that if several threads are accessing
* the variable at the same time, the logger needs to be thread-safe or it will produce rubbish.
*/
private Logger logger = MaryUtils.getLogger("DateEP");
public DateEP() {
super();
}
protected int canDealWith(String s, int type) {
return match(s, type);
}
protected int match(String s, int type) {
switch (type) {
case 0:
if (matchDateDMY(s)) return 1;
if (matchDateYMD(s)) return 2;
if (matchDateDM(s)) return 3;
// if (matchDateMY(s)) return 4;
// if (matchDateMDY(s)) return 8;
// if (matchDateMD(s)) return 9;
// if (matchDateM(s)) return 6;
// if (matchDateD(s)) return 7;
// if (matchDateY(s)) return 5;
break;
case 1:
if (matchDateDMY(s)) return 1;
break;
case 2:
if (matchDateYMD(s)) return 2;
break;
case 3:
if (matchDateDM(s)) return 3;
break;
case 4:
if (matchDateMY(s)) return 4;
break;
case 5:
if (matchDateY(s)) return 5;
break;
case 6:
if (matchDateM(s)) return 6;
break;
case 7:
if (matchDateD(s)) return 7;
break;
case 8:
if (matchDateMDY(s)) return 8;
break;
case 9:
if (matchDateMD(s)) return 9;
break;
}
return -1;
}
protected List<Element> expand(List<Element> tokens, String s, int type) {
if (tokens == null) throw new NullPointerException("Received null argument");
if (tokens.isEmpty()) throw new IllegalArgumentException("Received empty list");
Document doc = ((Element) tokens.get(0)).getOwnerDocument();
// we expect type to be one of the return values of match():
List<Element> expanded = null;
switch (type) {
case 1:
expanded = expandDateDMY(doc, s);
break;
case 2:
expanded = expandDateYMD(doc, s);
break;
case 3:
expanded = expandDateDM(doc, s);
break;
case 4:
expanded = expandDateMY(doc, s);
break;
case 5:
expanded = expandDateYear(doc, s);
break;
case 6:
expanded = expandDateMonth(doc, s);
break;
case 7:
expanded = expandDateDay(doc, s);
break;
case 8:
expanded = expandDateMDY(doc, s);
break;
case 9:
expanded = expandDateMD(doc, s);
break;
}
replaceTokens(tokens, expanded);
return expanded;
}
protected boolean matchDateDMY(String s) {
return reDayMonthYear.matcher(s).matches();
}
protected boolean matchDateYMD(String s) {
return reYearMonthDay.matcher(s).matches()
|| reDOWYearMonthDay.matcher(s).matches()
|| reYearMonthDayDOW.matcher(s).matches();
}
protected boolean matchDateDM(String s) {
return reDayMonth.matcher(s).matches();
}
protected boolean matchDateMY(String s) {
return reMonthYear.matcher(s).matches();
}
protected boolean matchDateY(String s) {
return reYear.matcher(s).matches();
}
protected boolean matchDateM(String s) {
return reMonth.matcher(s).matches();
}
protected boolean matchDateD(String s) {
return reDay.matcher(s).matches();
}
protected boolean matchDateMDY(String s) {
return reMonthDayYear.matcher(s).matches();
}
protected boolean matchDateMD(String s) {
return reMonthDay.matcher(s).matches();
}
protected List<Element> expandDateDMY(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reDayMonthYear.matcher(s);
boolean found = reMatcher.find();
if (!found) {
return null;
}
String day = reMatcher.group(1);
exp.addAll(number.expandCounted(doc, day + " 일", false));
String month = reMatcher.group(2);
exp.addAll(number.expandCounted(doc, month + " 월", false));
String year = reMatcher.group(3);
exp.addAll(number.expandCounted(doc, year + " 년", false));
return exp;
}
protected List<Element> expandDateYMD(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reDOWYearMonthDay.matcher(s);
boolean found = reMatcher.find();
int dayOfWeekType = 1;
if (!found) {
reMatcher = reYearMonthDayDOW.matcher(s);
found = reMatcher.find();
dayOfWeekType = 2; // month == Januar, Februar, ...
}
if (!found) {
reMatcher = reYearMonthDay.matcher(s);
found = reMatcher.find();
dayOfWeekType = 3; // month == Jan, Feb, ...
}
if (!found) {
return null;
}
String day = null, month = null, year = null, dow = null;
switch (dayOfWeekType) {
case 1:
dow = reMatcher.group(1);
year = reMatcher.group(2);
month = reMatcher.group(4);
day = reMatcher.group(5);
break;
case 2:
year = reMatcher.group(1);
month = reMatcher.group(3);
day = reMatcher.group(4);
dow = reMatcher.group(5);
break;
case 3:
year = reMatcher.group(1);
month = reMatcher.group(3);
day = reMatcher.group(4);
dow = null;
break;
}
if (dayOfWeekType == 1) {
exp.addAll(makeNewTokens(doc, dayOfWeekExpansion.get(dow)));
}
exp.addAll(number.expandCounted(doc, year + " 년", false));
exp.addAll(number.expandCounted(doc, month + " 월", false));
exp.addAll(number.expandCounted(doc, day + " 일", false));
if (dayOfWeekType == 2) {
exp.addAll(makeNewTokens(doc, dayOfWeekExpansion.get(dow)));
}
return exp;
}
protected List<Element> expandDateDM(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reDayMonth.matcher(s);
boolean found = reMatcher.find();
if (!found) {
return null;
}
String day = reMatcher.group(1);
exp.addAll(number.expandCounted(doc, day + " 일", false));
String month = reMatcher.group(2);
exp.addAll(number.expandCounted(doc, month + " 월", false));
return exp;
}
protected List<Element> expandDateMY(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reMonthYear.matcher(s);
boolean found = reMatcher.find();
if (!found) {
return null;
}
String year = reMatcher.group(3);
exp.addAll(number.expandCounted(doc, year + " 년", false));
String month = reMatcher.group(1);
if (month.charAt(0) == '0') month = month.substring(1);
exp.addAll(number.expandCounted(doc, month + " 월", false));
// alternatively: keep the number:
// exp.addAll(number.expandInteger(doc, month, false));
return exp;
}
protected List<Element> expandDateYear(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
exp.addAll(makeNewTokens(doc, number.expandInteger(s), true, s));
return exp;
}
protected List<Element> expandDateMonth(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reMonth.matcher(s);
boolean found = reMatcher.find();
if (!found) {
return null;
}
String month = reMatcher.group(1);
// Leading Zero must be deleted to get month from monthabbr.
if (month.charAt(0) == '0') month = month.substring(1);
exp.addAll(number.expandCounted(doc, month + " 월", false));
return exp;
}
protected List<Element> expandDateDay(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reDay.matcher(s);
boolean found = reMatcher.find();
if (!found) {
return null;
}
String day = reMatcher.group(1);
// Leading Zero must be deleted to get month from monthabbr.
if (day.charAt(0) == '0') day = day.substring(1);
exp.addAll(number.expandCounted(doc, day + " 일", false));
return exp;
}
protected List<Element> expandDateMD(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reMonthDay.matcher(s);
boolean found = reMatcher.find();
if (!found) {
return null;
}
String year = reMatcher.group(3);
exp.addAll(number.expandCounted(doc, year + " 년", false));
String month = reMatcher.group(1);
if (month.charAt(0) == '0') {
month = month.substring(1);
}
exp.addAll(number.expandCounted(doc, month + " 월", false));
// alternatively: keep the number:
// exp.addAll(number.expandInteger(doc, month, false));
String day;
day = reMatcher.group(2);
exp.addAll(number.expandCounted(doc, day + " 일", false));
return exp;
}
protected List<Element> expandDateMDY(Document doc, String s) {
ArrayList<Element> exp = new ArrayList<Element>();
Matcher reMatcher = reMonthDayYear.matcher(s);
boolean found = reMatcher.find();
if (!found) {
return null;
}
String day;
String year;
day = reMatcher.group(2);
year = reMatcher.group(3);
String month = reMatcher.group(1);
if (month.charAt(0) == '0') month = month.substring(1);
exp.addAll(number.expandCounted(doc, year + " 년", false));
exp.addAll(number.expandCounted(doc, month + " 월", false));
exp.addAll(number.expandCounted(doc, day + " 일", false));
return exp;
}
}
