private String getParentOrgan(
Document document,
Element description,
Element parentStructure,
String parentOfParentStructure) {
// check for 'part_of' relation on parentstructure
String partOfChain =
getPartOfChain(document, description, parentStructure, 0)
.replace(" of ", ",")
.trim(); // part of organ of organ
if (!partOfChain.isEmpty()) { // use explicit part_of
parentOfParentStructure = partOfChain;
}
// add constraint organ to parentorgan list
String parentStructureConstraint = parentStructure.getAttributeValue("constraint");
if (parentStructureConstraint != null) {
if (characterKnowledgeBase.isEntityStructuralContraint(parentStructureConstraint)) {
// parentorgan = constraint; //use the constraint of parentstruct as parentorgan, e.g. leaf
// blade ..., petiole ..., vein ....
parentOfParentStructure =
parentStructureConstraint + "," + parentOfParentStructure; // blade, leaf
}
}
// add name organ to parentorgan list
// parentorgan = parentofparentstructure+" "+parentstruct.getAttributeValue("name");//leaf blade
return parentStructure.getAttributeValue("name") + "," + parentOfParentStructure; // blade, leaf
}
/**
* @param struct
* @return true if the structure element has a constraint that refers to a(nother) structure
*/
private boolean hasStructuralConstraint(Element struct) {
String constraint = struct.getAttributeValue("constraint");
if (constraint == null || constraint.isEmpty()) return false;
if (characterKnowledgeBase.isEntityStructuralContraint(constraint)) return true;
return false;
}
@Override
public List<Token> tag(List<Token> sentence) {
List<Token> posedSentence = new ArrayList<Token>();
for (int i = 0; i < sentence.size(); i++) {
Token token = sentence.get(i);
String word = token.getContent();
String p = "";
boolean isState = false;
boolean isOrgan = false;
boolean isVerb = false;
// if(word.contains("~list~") || word.contains("_c_") ||
// organStateKnowledgeBase.isState(word))
// don't send numbers, values to learnedCharacterKnowledgeBase
if (word.contains("~list~")
|| word.contains("_c_")
|| ((word.matches("[^a-z]+")
|| word.contains("=")
|| word.matches(".*?(^|[^a-z])(" + units + ")([^a-z]|$).*"))
&& word.matches(".*?\\d.*"))) { // units could be mixed in the numbers
isState = true;
} else if (learnedCharacterKnowledgeBase.isCategoricalState(word)) {
isState = true;
CharacterMatch m = learnedCharacterKnowledgeBase.getCharacterName(word);
if (m != null
&& m.getCategories() != null
&& m.getCategories().matches("(^|.*?_)position_relational(_.*|$)")) {
isVerb = true;
}
} else {
// isOrgan = organStateKnowledgeBase.isOrgan(word);
isOrgan = learnedCharacterKnowledgeBase.isEntity(word);
}
Map<String, Set<String>> wordsToRoles = terminologyLearner.getWordsToRoles();
if (word.length() > 0
&& !word.matches("\\W")
&& !word.matches("(" + prepositions + ")")
&& !stopWords.contains(word)) {
if (wordsToRoles.containsKey(word)) p = wordsToRoles.get(word).iterator().next();
}
// Hong TODO modifiertokens
Matcher mc = compreppattern.matcher(word);
if (word.equals("lengths") || word.equals("length")) {
posedSentence.add(
new POSedToken(word, POS.NNS)); // stanford parser would sometimes take "lengths" as IN
} else if (word.equals("moreorless")) {
posedSentence.add(new POSedToken(word, POS.RB));
} else if (mc.matches()) {
posedSentence.add(new POSedToken(word, POS.IN));
} else if (word.matches("in-.*?(-view|profile)")) {
posedSentence.add(new POSedToken(word, POS.RB));
} else if (word.endsWith("ly") && word.indexOf("~") < 0) {
posedSentence.add(new POSedToken(word, POS.RB));
} else if (word.compareTo("becoming") == 0 || word.compareTo("about") == 0) {
posedSentence.add(new POSedToken(word, POS.RB));
} else if (word.compareTo("throughout") == 0
&& (i + 1 == sentence.size() || sentence.get(i + 1).getContent().matches("(,|or)"))) {
posedSentence.add(new POSedToken(word, POS.RB));
} else if (word.contains("#")) {
posedSentence.add(new POSedToken(word.replace('#', '-'), POS.RB));
} else if (word.compareTo("plus") == 0 || word.compareTo("and-or") == 0) {
posedSentence.add(new POSedToken(word, POS.CC));
} else if (word.matches("\\d+[cmd]?m\\d+[cmd]?m")) {
posedSentence.add(new POSedToken(word, POS.CD));
} else if (word.matches("(" + units + ")")) {
posedSentence.add(new POSedToken(word, POS.NNS));
} else if (word.matches("as-\\S+")) {
posedSentence.add(new POSedToken(word, POS.IN));
} else if (p.contains("op")) { // <inner> larger.
posedSentence.add(new POSedToken(word, POS.NNS));
} else if (p.contains("os") || (p.length() == 0 && isOrgan)) {
posedSentence.add(new POSedToken(word, POS.NN));
} else if (word.matches("(\\{?\\b" + roman + "\\b\\}?)")) { // mohan
// code
// to
// mark
// roman
// numbers
// {ii}
// or ii
// as
// ii/NNS
word = word.replaceAll("\\{|\\}", "");
posedSentence.add(new POSedToken(word, POS.NNS));
} else if (word.matches("\\d*.{0,1}\\d+")) {
posedSentence.add(new POSedToken(word, POS.CD));
} else if (isVerb) {
posedSentence.add(new POSedToken(word, POS.VB));
} else if (p.contains("c") || isState) {
int wordFrequency = corpus.getFrequency(word);
if (wordFrequency > 79) {
posedSentence.add(new Token(word));
} else {
posedSentence.add(new POSedToken(token.getContent(), POS.JJ));
}
} else {
posedSentence.add(new Token(token.getContent()));
}
}
return posedSentence;
}
