/**
* Convert an item in the Syllable relation into XML, inserting it at the specified location in
* the XML tree.
*
* @param deep whether to create a deep structure of <syllable> and <ph> elements or not.
*/
protected String insertSyllable(Item syllableItem, Element token, boolean deep) {
if (syllableItem == null || token == null) {
throw new NullPointerException("Null arguments to insertSyllable()");
}
if (!token.getTagName().equals(MaryXML.TOKEN)) {
throw new IllegalArgumentException("Syllables can only be inserted in <t> elements");
}
Document doc = token.getOwnerDocument();
Element syllable = null;
StringBuilder sampa = new StringBuilder();
if (deep) {
syllable = MaryXML.createElement(doc, MaryXML.SYLLABLE);
token.appendChild(syllable);
}
if (syllableItem.getFeatures().isPresent("accent")) {
String accentString = syllableItem.getFeatures().getString("accent");
if (deep) syllable.setAttribute("accent", accentString);
token.setAttribute("accent", accentString);
}
if (syllableItem.getFeatures().isPresent("stress")) {
String stressString = syllableItem.getFeatures().getString("stress");
if (!stressString.equals("0")) {
if (deep) syllable.setAttribute("stress", stressString);
if (stressString.equals("1")) sampa.append("'");
else if (stressString.equals("2")) sampa.append(",");
}
}
// Any segments?
Item segmentItem = syllableItem.getDaughter();
while (segmentItem != null) {
if (sampa.length() > 0) sampa.append(" ");
sampa.append(insertSegment(segmentItem, syllable, deep));
segmentItem = segmentItem.getNext();
}
String sampaString = sampa.toString();
if (deep) syllable.setAttribute("ph", sampaString);
// Any boundary?
if (syllableItem.getFeatures().isPresent("endtone")
&& !tokenItemHasFollowingBoundary(
syllableItem.getParent().getItemAs(Relation.TOKEN).getParent())) {
String endtone = syllableItem.getFeatures().getString("endtone");
if (!endtone.equals("")) {
Element boundary = MaryXML.createElement(doc, MaryXML.BOUNDARY);
boundary.setAttribute("tone", endtone);
boundary.setAttribute("breakindex", "4");
boundary.setAttribute("duration", "200");
// And the boundary comes after the current token:
token.getParentNode().appendChild(boundary);
}
}
return sampaString;
}
/**
* Convert an item in the Token relation into XML, inserting it at the specified location in the
* XML tree.
*
* @param deep whether to create a deep structure of <syllable> and <ph> elements or not.
*/
protected void insertToken(Item tokenItem, Element parent, boolean deep) {
if (tokenItem == null || parent == null) {
throw new NullPointerException("Null arguments to insertToken()");
}
Document doc = parent.getOwnerDocument();
Voice maryVoice = null;
if (tokenItem.getUtterance().getVoice() != null) {
maryVoice = FreeTTSVoices.getMaryVoice(tokenItem.getUtterance().getVoice());
}
AllophoneSet allophoneSet = (AllophoneSet) tokenItem.getUtterance().getObject("allophoneset");
if (allophoneSet == null) {
throw new NullPointerException(
"Utterance does not have an AllophoneSet -- should have been set in XML2UttBase.process()");
}
Element insertHere = parent;
boolean needMtu = false;
boolean insertPhonesFromToken = tokenItem.getFeatures().isPresent("phones");
Item testWordItem = null;
if (tokenItem.getFeatures().isPresent("precedingMarks")) {
String marks = tokenItem.getFeatures().getString("precedingMarks");
StringTokenizer markTok = new StringTokenizer(marks, ",");
while (markTok.hasMoreTokens()) {
String markStr = markTok.nextToken();
Element markEl = MaryXML.createElement(doc, MaryXML.MARK);
markEl.setAttribute("name", markStr);
insertHere.appendChild(markEl);
}
}
// Any boundary preceding the word?
if (tokenItem.getFeatures().isPresent("precedingBoundaryTone")
|| tokenItem.getFeatures().isPresent("precedingBoundaryBreakindex")
|| tokenItem.getFeatures().isPresent("precedingBoundaryDuration")) {
Element boundary = MaryXML.createElement(doc, MaryXML.BOUNDARY);
insertHere.appendChild(boundary);
if (tokenItem.getFeatures().isPresent("precedingBoundaryTone"))
boundary.setAttribute("tone", tokenItem.getFeatures().getString("precedingBoundaryTone"));
if (tokenItem.getFeatures().isPresent("precedingBoundaryBreakindex"))
boundary.setAttribute(
"breakindex", tokenItem.getFeatures().getString("precedingBoundaryBreakindex"));
if (tokenItem.getFeatures().isPresent("precedingBoundaryDuration"))
boundary.setAttribute(
"duration", tokenItem.getFeatures().getString("precedingBoundaryDuration"));
}
if (tokenItem.getNthDaughter(1) != null
|| (testWordItem = tokenItem.getDaughter()) != null
&& !testWordItem.toString().equals(tokenItem.toString().toLowerCase())) {
// Token has more than one daughter, or the only daughter is an
// expanded form -- need to create an <mtu> element
needMtu = true;
Element mtu = MaryXML.createElement(doc, MaryXML.MTU);
parent.appendChild(mtu);
mtu.setAttribute("orig", tokenItem.toString());
insertHere = mtu;
}
// Any words?
FeatureSet tokenFeatureSet = tokenItem.getFeatures();
Item tokenDaughter = tokenItem.getDaughter();
if (tokenDaughter == null) { // no word relation present
// Create a <t> element based on token information only
Element t = MaryXML.createElement(doc, MaryXML.TOKEN);
insertHere.appendChild(t);
MaryDomUtils.setTokenText(t, tokenItem.toString());
if (insertPhonesFromToken) {
String[] phones = (String[]) tokenItem.getFeatures().getObject("phones");
t.setAttribute("ph", phoneArray2phoneString(allophoneSet, phones));
insertPhonesFromToken = false;
}
if (tokenFeatureSet.isPresent("accent")) {
t.setAttribute("accent", tokenFeatureSet.getString("accent"));
}
}
while (tokenDaughter != null) {
// Part of speech, if present, is associated with the word
// relation.
Item wordItem = tokenDaughter.getItemAs("Word");
Element t = null;
StringBuilder sampa = new StringBuilder();
if (wordItem != null) {
t = MaryXML.createElement(doc, MaryXML.TOKEN);
insertHere.appendChild(t);
String tokenText = null;
// If there is only one, non-expanded word, use text from
// tokenItem in order to retain capitalisation:
if (needMtu) tokenText = wordItem.toString();
else tokenText = tokenItem.toString();
MaryDomUtils.setTokenText(t, tokenText);
if (insertPhonesFromToken) {
String[] phones = (String[]) tokenItem.getFeatures().getObject("phones");
t.setAttribute("ph", phoneArray2phoneString(allophoneSet, phones));
insertPhonesFromToken = false;
} else if (wordItem.getFeatures().isPresent("phones")) {
// the word item has phones, take them only if there are no Token phones
String[] phones = (String[]) wordItem.getFeatures().getObject("phones");
t.setAttribute("ph", phoneArray2phoneString(allophoneSet, phones));
}
if (tokenFeatureSet.isPresent("accent")) {
t.setAttribute("accent", tokenFeatureSet.getString("accent"));
}
FeatureSet wordFeatureSet = wordItem.getFeatures();
if (wordFeatureSet.isPresent("pos")) t.setAttribute("pos", wordFeatureSet.getString("pos"));
}
// Any syllables?
Item sylStruct = tokenDaughter.getItemAs("SylStructure");
if (sylStruct != null && sylStruct.hasDaughters()) {
Item syllableItem = sylStruct.getDaughter();
while (syllableItem != null) {
if (sampa.length() > 0) sampa.append(" - ");
sampa.append(insertSyllable(syllableItem, t, deep));
syllableItem = syllableItem.getNext();
}
}
if (sampa.length() > 0) t.setAttribute("ph", sampa.toString());
tokenDaughter = tokenDaughter.getNext();
}
// Any marks after the word but before the punctuation?
if (tokenItem.getFeatures().isPresent("prePuncMarks")) {
String marks = tokenItem.getFeatures().getString("prePuncMarks");
StringTokenizer markTok = new StringTokenizer(marks, ",");
while (markTok.hasMoreTokens()) {
String markStr = markTok.nextToken();
Element markEl = MaryXML.createElement(doc, MaryXML.MARK);
markEl.setAttribute("name", markStr);
insertHere.appendChild(markEl);
}
}
// Any punctuation after the word?
if (tokenItem.getFeatures().isPresent("punc")) {
String puncString = tokenItem.getFeatures().getString("punc");
if (!puncString.equals("")) {
Element punctuation = MaryXML.createElement(doc, MaryXML.TOKEN);
MaryDomUtils.setTokenText(punctuation, puncString);
String pos = null;
if (puncString.equals(",")) pos = "$,";
else pos = "$PUNCT";
punctuation.setAttribute("pos", pos);
parent.appendChild(punctuation);
}
}
// Any marks after the word?
if (tokenItem.getFeatures().isPresent("followingMarks")) {
String marks = tokenItem.getFeatures().getString("followingMarks");
StringTokenizer markTok = new StringTokenizer(marks, ",");
while (markTok.hasMoreTokens()) {
String markStr = markTok.nextToken();
Element markEl = MaryXML.createElement(doc, MaryXML.MARK);
markEl.setAttribute("name", markStr);
insertHere.appendChild(markEl);
}
}
// Any boundary after the word?
if (tokenItemHasFollowingBoundary(tokenItem)) {
Element boundary = MaryXML.createElement(doc, MaryXML.BOUNDARY);
insertHere.appendChild(boundary);
if (tokenItem.getFeatures().isPresent("followingBoundaryTone"))
boundary.setAttribute("tone", tokenItem.getFeatures().getString("followingBoundaryTone"));
int breakindex = 0;
if (tokenItem.getFeatures().isPresent("followingBoundaryBreakindex")) {
String breakindexString = tokenItem.getFeatures().getString("followingBoundaryBreakindex");
boundary.setAttribute("breakindex", breakindexString);
try {
breakindex = Integer.parseInt(breakindexString);
} catch (NumberFormatException nfe) {
}
}
if (tokenItem.getFeatures().isPresent("followingBoundaryDuration"))
boundary.setAttribute(
"duration", tokenItem.getFeatures().getString("followingBoundaryDuration"));
else { // estimate reasonable duration values based on the break index
if (breakindex >= 4) {
boundary.setAttribute("duration", "400");
} else if (breakindex == 3) {
boundary.setAttribute("duration", "200");
} // and no duration for boundaries with bi < 3
}
}
}
/**
* Depending on the data type, find the right information in the utterance and insert it into the
* sentence.
*/
protected final void fillSentence(Element sentence, Utterance utterance) {
Document doc = sentence.getOwnerDocument();
Relation tokenRelation = utterance.getRelation(Relation.TOKEN);
if (tokenRelation == null) return;
Item tokenItem = tokenRelation.getHead();
Relation phraseRelation = utterance.getRelation(Relation.PHRASE);
Item phraseItem = null;
if (phraseRelation != null) {
phraseItem = phraseRelation.getHead();
// Challenge: Bring token and phrase relations together. They have
// common children, which can be interpreted as Word or SylStructure
// items. Algorithm: For a given phrase, look at tokens. If a token's
// first child, interpreted in the phrase relation, has the phrase as
// its parent, then insert the token and all its children, and move to
// the next token. If not, move to the next phrase.
while (phraseItem != null) {
// The phrases:
Element phrase = MaryXML.createElement(doc, MaryXML.PHRASE);
sentence.appendChild(phrase);
Element insertHere = phrase;
// Is this token part of this phrase?
while (tokenItem != null
&& tokenItem.getDaughter().findItem("R:Phrase.parent").equals(phraseItem)) {
FeatureSet tokenFeatures = tokenItem.getFeatures();
if (tokenFeatures.isPresent(XML2UttBase.PROSODY_START)) {
Element prosody = insertProsodySettings(insertHere, tokenFeatures);
if (prosody != null) {
insertHere = prosody;
}
}
insertToken(tokenItem, phrase, true); // create deep structure
if (tokenFeatures.isPresent(XML2UttBase.PROSODY_END)) {
assert insertHere.getTagName().equals(MaryXML.PROSODY);
insertHere = (Element) insertHere.getParentNode();
}
tokenItem = tokenItem.getNext();
}
phraseItem = phraseItem.getNext();
}
} else {
// No phrase relation, simply create tokens.
Element insertHere = sentence;
while (tokenItem != null) {
FeatureSet tokenFeatures = tokenItem.getFeatures();
if (tokenFeatures.isPresent(XML2UttBase.PROSODY_START)) {
Element prosody = insertProsodySettings(insertHere, tokenFeatures);
if (prosody != null) {
insertHere = prosody;
}
}
insertToken(tokenItem, insertHere);
if (tokenFeatures.isPresent(XML2UttBase.PROSODY_END)) {
if (insertHere.getTagName().equals(MaryXML.PROSODY)) {
insertHere = (Element) insertHere.getParentNode();
} // else, we are looking at an empty prosody tag with no arguments, which is being
// deleted right now.
}
tokenItem = tokenItem.getNext();
}
}
}