/**
* Convert an item in the Segment 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 insertSegment(Item segmentItem, Element syllable, boolean deep) {
// allow for syllable == null if not deep:
if (segmentItem == null || deep && syllable == null) {
throw new NullPointerException("Null arguments to insertSegment()");
}
if (deep && !syllable.getTagName().equals(MaryXML.SYLLABLE)) {
throw new IllegalArgumentException("Segments can only be inserted in <syllable> elements");
}
String segmentString = segmentItem.toString();
Voice maryVoice = FreeTTSVoices.getMaryVoice(segmentItem.getUtterance().getVoice());
if (deep) {
Document doc = syllable.getOwnerDocument();
Element segment = MaryXML.createElement(doc, MaryXML.PHONE);
syllable.appendChild(segment);
segment.setAttribute("p", segmentString);
if (segmentItem.getFeatures().isPresent("end")) {
float endInSeconds = segmentItem.getFeatures().getFloat("end");
int endInMillis = (int) (1000 * endInSeconds);
segment.setAttribute("end", String.valueOf(endInMillis));
}
if (segmentItem.getFeatures().isPresent("mbr_dur")) {
int mbrDur = segmentItem.getFeatures().getInt("mbr_dur");
segment.setAttribute("d", String.valueOf(mbrDur));
}
if (segmentItem.getFeatures().isPresent("mbr_targets")) {
String mbrTargets = segmentItem.getFeatures().getString("mbr_targets");
if (!mbrTargets.equals("")) {
segment.setAttribute("f0", mbrTargets);
}
}
}
return segmentString;
}
/**
* Changes the pronunciation of "the" from 'd ax' to 'd iy' if the following word starts with a
* vowel. "The every" is a good example.
*
* @param utterance the utterance to process
*/
private void fixTheIy(Utterance utterance) {
Voice voice = utterance.getVoice();
for (Item item = utterance.getRelation(Relation.SEGMENT).getHead();
item != null;
item = item.getNext()) {
if ("ax".equals(item.toString())) {
String word = wordPath.findFeature(item).toString();
if ("the".equals(word) && ("+".equals(N_PH_VC.findFeature(item)))) {
item.getFeatures().setString("name", "iy");
}
}
}
}
/**
* 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
}
}
}