Package edu.stanford.nlp.parser.lexparser

Source Code of edu.stanford.nlp.parser.lexparser.ChineseTreebankParserParams

package edu.stanford.nlp.parser.lexparser;

import edu.stanford.nlp.io.EncodingPrintWriter;
import edu.stanford.nlp.io.NumberRangesFileFilter;
import edu.stanford.nlp.io.RuntimeIOException;
import edu.stanford.nlp.ling.CategoryWordTag;
import edu.stanford.nlp.ling.CoreLabel;
import edu.stanford.nlp.ling.Label;
import edu.stanford.nlp.ling.Sentence;
import edu.stanford.nlp.ling.Word;
import edu.stanford.nlp.process.WordSegmenter;
import edu.stanford.nlp.process.WordSegmentingTokenizer;
import edu.stanford.nlp.trees.*;
import edu.stanford.nlp.trees.international.pennchinese.*;
import java.util.function.Predicate;
import edu.stanford.nlp.util.Index;
import edu.stanford.nlp.util.ReflectionLoading;
import edu.stanford.nlp.util.StringUtils;

import java.io.IOException;
import java.util.*;


/**
* Parameter file for parsing the Penn Chinese Treebank.  Includes
* category enrichments specific to the Penn Chinese Treebank.
*
* @author Roger Levy
* @author Christopher Manning
* @author Galen Andrew
*/

public class ChineseTreebankParserParams extends AbstractTreebankParserParams {

  /**
   * The variable ctlp stores the same thing as the tlp variable in
   * AbstractTreebankParserParams, but pre-cast to be a
   * ChineseTreebankLanguagePack.
   * todo [cdm 2013]: Just change to method that casts
   */
  private ChineseTreebankLanguagePack ctlp;
  public boolean charTags = false;
  public boolean useCharacterBasedLexicon = false;
  public boolean useMaxentLexicon = false;
  public boolean useMaxentDepGrammar = false;
  public boolean segment = false;
  public boolean segmentMarkov = false;
  public boolean sunJurafskyHeadFinder = false;
  public boolean bikelHeadFinder = false;
  public boolean discardFrags = false;
  public boolean useSimilarWordMap = false;

  public String segmenterClass = null;

  private Lexicon lex;
  private WordSegmenter segmenter;
  private HeadFinder headFinder = null;

  private static void printlnErr(String s) {
    EncodingPrintWriter.err.println(s, ChineseTreebankLanguagePack.ENCODING);
  }

  public ChineseTreebankParserParams() {
    super(new ChineseTreebankLanguagePack());
    ctlp = (ChineseTreebankLanguagePack) super.treebankLanguagePack();
  }

  /**
   * Returns a ChineseHeadFinder
   */
  @Override
  public HeadFinder headFinder() {
    if(headFinder == null) {
      if (sunJurafskyHeadFinder) {
        return new SunJurafskyChineseHeadFinder();
      } else if (bikelHeadFinder) {
        return new BikelChineseHeadFinder();
      } else {
        return new ChineseHeadFinder();
      }
    } else
      return headFinder;
  }

  @Override
  public HeadFinder typedDependencyHeadFinder() {
    return new ChineseSemanticHeadFinder();
  }

  /**
   * Returns a ChineseLexicon
   */
  @Override
  public Lexicon lex(Options op, Index<String> wordIndex, Index<String> tagIndex) {
    if (useCharacterBasedLexicon) {
      return lex = new ChineseCharacterBasedLexicon(this, wordIndex, tagIndex);
    // } else if (useMaxentLexicon) {
    // return lex = new ChineseMaxentLexicon();
    }
    if (op.lexOptions.uwModelTrainer == null) {
      op.lexOptions.uwModelTrainer = "edu.stanford.nlp.parser.lexparser.ChineseUnknownWordModelTrainer";
    }
    if (segmenterClass != null) {
      try {
        segmenter = ReflectionLoading.loadByReflection(segmenterClass, this,
                                                       wordIndex, tagIndex);
      } catch (ReflectionLoading.ReflectionLoadingException e) {
        segmenter = ReflectionLoading.loadByReflection(segmenterClass);
      }
    }

    ChineseLexicon clex = new ChineseLexicon(op, this, wordIndex, tagIndex);
    if (segmenter != null) {
      lex = new ChineseLexiconAndWordSegmenter(clex, segmenter);
      ctlp.setTokenizerFactory(WordSegmentingTokenizer.factory(segmenter));
    } else {
      lex = clex;
    }

    return lex;
  }

  @Override
  public double[] MLEDependencyGrammarSmoothingParams() {
    return new double[]{5.8, 17.7, 6.5, 0.4};
  }

  @Override
  public TreeReaderFactory treeReaderFactory() {
    final TreeNormalizer tn = new CTBErrorCorrectingTreeNormalizer(splitNPTMP, splitPPTMP, splitXPTMP, charTags);
    return new CTBTreeReaderFactory(tn, discardFrags);
  }

  /**
   * Uses a DiskTreebank with a CHTBTokenizer and a
   * BobChrisTreeNormalizer.
   */
  @Override
  public DiskTreebank diskTreebank() {
    String encoding = inputEncoding;
    if (!java.nio.charset.Charset.isSupported(encoding)) {
      printlnErr("Warning: desired encoding " + encoding + " not accepted. ");
      printlnErr("Using UTF-8 to construct DiskTreebank");
      encoding = "UTF-8";
    }

    return new DiskTreebank(treeReaderFactory(), encoding);
  }


  /**
   * Uses a MemoryTreebank with a CHTBTokenizer and a
   * BobChrisTreeNormalizer
   */
  @Override
  public MemoryTreebank memoryTreebank() {
    String encoding = inputEncoding;
    if (!java.nio.charset.Charset.isSupported(encoding)) {
      System.out.println("Warning: desired encoding " + encoding + " not accepted. ");
      System.out.println("Using UTF-8 to construct MemoryTreebank");
      encoding = "UTF-8";
    }

    return new MemoryTreebank(treeReaderFactory(), encoding);
  }


  /**
   * Returns a ChineseCollinizer
   */
  @Override
  public TreeTransformer collinizer() {
    return new ChineseCollinizer(ctlp);
  }

  /**
   * Returns a ChineseCollinizer that doesn't delete punctuation
   */
  @Override
  public TreeTransformer collinizerEvalb() {
    return new ChineseCollinizer(ctlp, false);
  }

  //   /** Returns a <code>ChineseTreebankLanguagePack</code> */
  //   public TreebankLanguagePack treebankLanguagePack() {
  //     return new ChineseTreebankLanguagePack();
  //   }


  /* --------- not used now
    // Automatically generated by ParentAnnotationStats -- preferably don't edit
    private static final String[] splitters1 = new String[] {"VA^VCD", "NP^NP", "NP^VP", "NP^IP", "NP^DNP", "NP^PP", "NP^LCP", "NP^PRN", "NP^QP", "PP^IP", "PP^NP", "NN^FRAG", "NN^NP", "NT^FRAG", "NT^NP", "NR^FRAG", "NR^NP", "VV^FRAG", "VV^VRD", "VV^VCD", "VV^VP", "VV^VSB", "VP^VP", "VP^IP", "VP^DVP", "IP^ROOT", "IP^IP", "IP^CP", "IP^VP", "IP^PP", "IP^NP", "IP^LCP", "CP^IP", "QP^NP", "QP^PP", "QP^VP", "ADVP^CP", "CC^VP", "CC^NP", "CC^IP", "CC^QP", "PU^NP", "PU^FRAG", "PU^IP", "PU^VP", "PU^PRN", "PU^QP", "PU^LST", "NP^DNP~QP", "NT^NP~NP", "NT^NP~VP", "NT^NP~IP", "NT^NP~LCP", "NT^NP~PP", "NT^NP~PRN", "NT^NP~QP", "NT^NP~DNP", "NP^NP~VP", "NP^NP~NP", "NP^NP~IP", "NP^NP~PP", "NP^NP~DNP", "NP^NP~LCP", "NN^NP~VP", "NN^NP~IP", "NN^NP~NP", "NN^NP~PP", "NN^NP~DNP", "NN^NP~LCP", "NN^NP~UCP", "NN^NP~QP", "NN^NP~PRN", "M^CLP~DP", "M^CLP~QP", "M^CLP~NP", "M^CLP~CLP", "CD^QP~VP", "CD^QP~NP", "CD^QP~QP", "CD^QP~LCP", "CD^QP~PP", "CD^QP~DNP", "CD^QP~DP", "CD^QP~IP", "IP^IP~IP", "IP^IP~ROOT", "IP^IP~VP", "LC^LCP~PP", "LC^LCP~IP", "NP^VP~IP", "NP^VP~VP", "AD^ADVP~IP", "AD^ADVP~QP", "AD^ADVP~VP", "AD^ADVP~NP", "AD^ADVP~PP", "AD^ADVP~ADVP", "NP^IP~ROOT", "NP^IP~IP", "NP^IP~CP", "NP^IP~VP", "DT^DP~PP", "P^PP~IP", "P^PP~NP", "P^PP~VP", "P^PP~DNP", "VV^VP~IP", "VV^VP~VP", "PU^IP~IP", "PU^IP~VP", "PU^IP~ROOT", "PU^IP~CP", "JJ^ADJP~DNP", "JJ^ADJP~ADJP", "NR^NP~IP", "NR^NP~NP", "NR^NP~PP", "NR^NP~VP", "NR^NP~DNP", "NR^NP~LCP", "NR^NP~PRN", "NP^PP~NP", "NP^PP~IP", "NP^PP~DNP", "VA^VP~VP", "VA^VP~IP", "VA^VP~DVP", "VP^VP~VP", "VP^VP~IP", "VP^VP~DVP", "VP^IP~ROOT", "VP^IP~CP", "VP^IP~IP", "VP^IP~VP", "VP^IP~PP", "VP^IP~LCP", "VP^IP~NP", "PN^NP~NP", "PN^NP~IP", "PN^NP~PP"};
    private static final String[] splitters2 = new String[] {"VA^VCD", "NP^NP", "NP^VP", "NP^IP", "NP^DNP", "NP^PP", "NP^LCP", "NN^FRAG", "NN^NP", "NT^FRAG", "NT^NP", "NR^FRAG", "NR^NP", "VV^FRAG", "VV^VRD", "VV^VCD", "VV^VP", "VV^VSB", "VP^VP", "VP^IP", "VP^DVP", "IP^ROOT", "IP^IP", "IP^CP", "IP^VP", "IP^PP", "CP^IP", "ADVP^CP", "CC^VP", "CC^NP", "PU^NP", "PU^FRAG", "PU^IP", "PU^VP", "PU^PRN", "NT^NP~NP", "NT^NP~VP", "NT^NP~IP", "NT^NP~LCP", "NT^NP~PP", "NP^NP~VP", "NP^NP~NP", "NP^NP~IP", "NP^NP~PP", "NP^NP~DNP", "NN^NP~VP", "NN^NP~IP", "NN^NP~NP", "NN^NP~PP", "NN^NP~DNP", "NN^NP~LCP", "NN^NP~UCP", "NN^NP~QP", "NN^NP~PRN", "M^CLP~DP", "CD^QP~VP", "CD^QP~NP", "CD^QP~QP", "CD^QP~LCP", "CD^QP~PP", "CD^QP~DNP", "CD^QP~DP", "LC^LCP~PP", "NP^VP~IP", "NP^VP~VP", "AD^ADVP~IP", "AD^ADVP~QP", "AD^ADVP~VP", "AD^ADVP~NP", "NP^IP~ROOT", "NP^IP~IP", "NP^IP~CP", "NP^IP~VP", "P^PP~IP", "P^PP~NP", "P^PP~VP", "P^PP~DNP", "VV^VP~IP", "VV^VP~VP", "PU^IP~IP", "PU^IP~VP", "PU^IP~ROOT", "PU^IP~CP", "JJ^ADJP~DNP", "NR^NP~IP", "NR^NP~NP", "NR^NP~PP", "NR^NP~VP", "NR^NP~DNP", "NR^NP~LCP", "NP^PP~NP", "VA^VP~VP", "VA^VP~IP", "VP^VP~VP", "VP^IP~ROOT", "VP^IP~CP", "VP^IP~IP", "VP^IP~VP", "VP^IP~PP", "VP^IP~LCP", "VP^IP~NP", "PN^NP~NP"};
    private static final String[] splitters3 = new String[] {"NP^NP", "NP^VP", "NP^IP", "NP^DNP", "NP^PP", "NP^LCP", "NN^FRAG", "NN^NP", "NT^FRAG", "NR^FRAG", "NR^NP", "VV^FRAG", "VV^VRD", "VV^VCD", "VV^VP", "VV^VSB", "VP^VP", "VP^IP", "IP^ROOT", "IP^IP", "IP^CP", "IP^VP", "PU^NP", "PU^FRAG", "PU^IP", "PU^VP", "PU^PRN", "NP^NP~VP", "NN^NP~VP", "NN^NP~IP", "NN^NP~NP", "NN^NP~PP", "NN^NP~DNP", "NN^NP~LCP", "M^CLP~DP", "CD^QP~VP", "CD^QP~NP", "CD^QP~QP", "AD^ADVP~IP", "AD^ADVP~QP", "AD^ADVP~VP", "P^PP~IP", "VV^VP~IP", "VV^VP~VP", "PU^IP~IP", "PU^IP~VP", "NR^NP~IP", "NR^NP~NP", "NR^NP~PP", "NR^NP~VP", "VP^VP~VP", "VP^IP~ROOT", "VP^IP~CP", "VP^IP~IP", "VP^IP~VP"};
    private static final String[] splitters4 = new String[] {"NP^NP", "NP^VP", "NP^IP", "NN^FRAG", "NT^FRAG", "NR^FRAG", "VV^FRAG", "VV^VRD", "VV^VCD", "VP^VP", "VP^IP", "IP^ROOT", "IP^IP", "IP^CP", "IP^VP", "PU^NP", "PU^FRAG", "PU^IP", "PU^VP", "NN^NP~VP", "NN^NP~IP", "NN^NP~NP", "NN^NP~PP", "NN^NP~DNP", "NN^NP~LCP", "CD^QP~VP", "CD^QP~NP", "AD^ADVP~IP", "VV^VP~IP", "VV^VP~VP", "NR^NP~IP", "VP^IP~ROOT", "VP^IP~CP"};
    // these ones were built by hand.
    // one can't tag split under FRAG or everything breaks, because of those
    // big flat FRAGs....
    private static final String[] splitters5 = new String[] {"NN^FRAG", "NT^FRAG", "NR^FRAG", "VV^FRAG", "VV^VCD", "VV^VRD", "NP^NP", "VP^VP", "IP^ROOT", "IP^IP", "PU^NP", "PU^FRAG", "P^PP~VP", "P^PP~IP"};
    private static final String[] splitters6 = new String[] {"VV^VCD", "VV^VRD", "NP^NP", "VP^VP", "IP^ROOT", "IP^IP", "PU^NP", "P^PP~VP", "P^PP~IP"};
    private static final String[] splitters7 = new String[] {"NP^NP", "VP^VP", "IP^ROOT", "IP^IP", "PU^NP", "P^PP~VP", "P^PP~IP"};
    private static final String[] splitters8 = new String[] {"IP^ROOT", "IP^IP", "PU^NP", "P^PP~VP", "P^PP~IP"};
    private static final String[] splitters9 = new String[] {"VV^VCD", "VV^VRD", "NP^NP", "VP^VP", "IP^ROOT", "IP^IP", "P^PP~VP", "P^PP~IP"};
    private static final String[] splitters10 = new String[] {"NP^NP", "VP^VP", "IP^ROOT", "IP^IP", "P^PP~VP", "P^PP~IP"};


    public String[] splitters() {
      switch (selectiveSplitLevel) {
      case 1:
        return splitters1;
      case 2:
        return splitters2;
      case 3:
        return splitters3;
      case 4:
        return splitters4;
      case 5:
        return splitters5;
      case 6:
        return splitters6;
      case 7:
        return splitters7;
      case 8:
        return splitters8;
      case 9:
        return splitters9;
      case 10:
        return splitters10;
      default:
        return new String[0];
      }
    }
  ------------------ */

  @Override
  public String[] sisterSplitters() {
    return StringUtils.EMPTY_STRING_ARRAY;
  }

  /**
   * transformTree does all language-specific tree
   * transformations. Any parameterizations should be inside the
   * specific TreebankLangParserParams class.
   */
  @Override
  public Tree transformTree(Tree t, Tree root) {
    if (t == null || t.isLeaf()) {
      return t;
    }

    String parentStr;
    String grandParentStr;
    Tree parent;
    Tree grandParent;
    if (root == null || t.equals(root)) {
      parent = null;
      parentStr = "";
    } else {
      parent = t.parent(root);
      parentStr = parent.label().value();
    }
    if (parent == null || parent.equals(root)) {
      grandParent = null;
      grandParentStr = "";
    } else {
      grandParent = parent.parent(root);
      grandParentStr = grandParent.label().value();
    }

    String baseParentStr = ctlp.basicCategory(parentStr);
    String baseGrandParentStr = ctlp.basicCategory(grandParentStr);

    CoreLabel lab = (CoreLabel) t.label();
    String word = lab.word();
    String tag = lab.tag();
    String baseTag = ctlp.basicCategory(tag);
    String category = lab.value();
    String baseCategory = ctlp.basicCategory(category);

    if (t.isPreTerminal()) { // it's a POS tag
      List<String> leftAunts = listBasicCategories(SisterAnnotationStats.leftSisterLabels(parent, grandParent));
      List<String> rightAunts = listBasicCategories(SisterAnnotationStats.rightSisterLabels(parent, grandParent));

      // Chinese-specific punctuation splits
      if (chineseSplitPunct && baseTag.equals("PU")) {
        if (ChineseTreebankLanguagePack.chineseDouHaoAcceptFilter().test(word)) {
          tag = tag + "-DOU";
          // System.out.println("Punct: Split dou hao"); // debugging
        } else if (ChineseTreebankLanguagePack.chineseCommaAcceptFilter().test(word)) {
          tag = tag + "-COMMA";
          // System.out.println("Punct: Split comma"); // debugging
        } else if (ChineseTreebankLanguagePack.chineseColonAcceptFilter().test(word)) {
          tag = tag + "-COLON";
          // System.out.println("Punct: Split colon"); // debugging
        } else if (ChineseTreebankLanguagePack.chineseQuoteMarkAcceptFilter().test(word)) {
          if (chineseSplitPunctLR) {
            if (ChineseTreebankLanguagePack.chineseLeftQuoteMarkAcceptFilter().test(word)) {
              tag += "-LQUOTE";
            } else {
              tag += "-RQUOTE";
            }
          } else {
            tag = tag + "-QUOTE";
          }
          // System.out.println("Punct: Split quote"); // debugging
        } else if (ChineseTreebankLanguagePack.chineseEndSentenceAcceptFilter().test(word)) {
          tag = tag + "-ENDSENT";
          // System.out.println("Punct: Split end sent"); // debugging
        } else if (ChineseTreebankLanguagePack.chineseParenthesisAcceptFilter().test(word)) {
          if (chineseSplitPunctLR) {
            if (ChineseTreebankLanguagePack.chineseLeftParenthesisAcceptFilter().test(word)) {
              tag += "-LPAREN";
            } else {
              tag += "-RPAREN";
            }
          } else {
            tag += "-PAREN";
            //printlnErr("Just used -PAREN annotation");
            //printlnErr(word);
            //throw new RuntimeException();
          }
          // System.out.println("Punct: Split paren"); // debugging
        } else if (ChineseTreebankLanguagePack.chineseDashAcceptFilter().test(word)) {
          tag = tag + "-DASH";
          // System.out.println("Punct: Split dash"); // debugging
        } else if (ChineseTreebankLanguagePack.chineseOtherAcceptFilter().test(word)) {
          tag = tag + "-OTHER";
        } else {
          printlnErr("Unknown punct (you should add it to CTLP): " + tag + " |" + word + "|");
        }
      } else if (chineseSplitDouHao) {   // only split DouHao
        if (ChineseTreebankLanguagePack.chineseDouHaoAcceptFilter().test(word) && baseTag.equals("PU")) {
          tag = tag + "-DOU";
        }
      }

      // Chinese-specific POS tag splits (non-punctuation)

      if (tagWordSize) {
        int l = word.length();
        tag += "-" + l + "CHARS";
      }

      if (mergeNNVV && baseTag.equals("NN")) {
        tag = "VV";
      }

      if ((chineseSelectiveTagPA || chineseVerySelectiveTagPA) && (baseTag.equals("CC") || baseTag.equals("P"))) {
        tag += "-" + baseParentStr;
      }
      if (chineseSelectiveTagPA && (baseTag.equals("VV"))) {
        tag += "-" + baseParentStr;
      }

      if (markMultiNtag && tag.startsWith("N")) {
        for (int i = 0; i < parent.numChildren(); i++) {
          if (parent.children()[i].label().value().startsWith("N") && parent.children()[i] != t) {
            tag += "=N";
            //System.out.println("Found multi=N rewrite");
          }
        }
      }

      if (markVVsisterIP && baseTag.equals("VV")) {
        boolean seenIP = false;
        for (int i = 0; i < parent.numChildren(); i++) {
          if (parent.children()[i].label().value().startsWith("IP")) {
            seenIP = true;
          }
        }
        if (seenIP) {
          tag += "-IP";
          //System.out.println("Found VV with IP sister"); // testing
        }
      }

      if (markPsisterIP && baseTag.equals("P")) {
        boolean seenIP = false;
        for (int i = 0; i < parent.numChildren(); i++) {
          if (parent.children()[i].label().value().startsWith("IP")) {
            seenIP = true;
          }
        }
        if (seenIP) {
          tag += "-IP";
        }
      }

      if (markADgrandchildOfIP && baseTag.equals("AD") && baseGrandParentStr.equals("IP")) {
        tag += "~IP";
        //System.out.println("Found AD with IP grandparent"); // testing
      }

      if (gpaAD && baseTag.equals("AD")) {
        tag += "~" + baseGrandParentStr;
        //System.out.println("Found AD with grandparent " + grandParentStr); // testing
      }

      if (markPostverbalP && leftAunts.contains("VV") && baseTag.equals("P")) {
        //System.out.println("Found post-verbal P");
        tag += "^=lVV";
      }

      // end Chinese-specific tag splits

      Label label = new CategoryWordTag(tag, word, tag);
      t.setLabel(label);
    } else {
      // it's a phrasal category
      Tree[] kids = t.children();

      // Chinese-specific category splits
      List<String> leftSis = listBasicCategories(SisterAnnotationStats.leftSisterLabels(t, parent));
      List<String> rightSis = listBasicCategories(SisterAnnotationStats.rightSisterLabels(t, parent));

      if (paRootDtr && baseParentStr.equals("ROOT")) {
        category += "^ROOT";
      }

      if (markIPsisterBA && baseCategory.equals("IP")) {
        if (leftSis.contains("BA")) {
          category += "=BA";
          //System.out.println("Found IP sister of BA");
        }
      }

      if (dominatesV && hasV(t.preTerminalYield())) {
        // mark categories containing a verb
        category += "-v";
      }

      if (markIPsisterVVorP && baseCategory.equals("IP")) {
        // todo: cdm: is just looking for "P" here selective enough??
        if (leftSis.contains("VV") || leftSis.contains("P")) {
          category += "=VVP";
        }
      }

      if (markIPsisDEC && baseCategory.equals("IP")) {
        if (rightSis.contains("DEC")) {
          category += "=DEC";
          //System.out.println("Found prenominal IP");
        }
      }

      if (baseCategory.equals("VP")) {
        // cdm 2008: this used to just check that it startsWith("VP"), but
        // I think that was bad because it also matched VPT verb compounds
        if (chineseSplitVP == 3) {
          boolean hasCC = false;
          boolean hasPU = false;
          boolean hasLexV = false;
          for (Tree kid : kids) {
            if (kid.label().value().startsWith("CC")) {
              hasCC = true;
            } else if (kid.label().value().startsWith("PU")) {
              hasPU = true;
            } else if (StringUtils.lookingAt(kid.label().value(), "(V[ACEV]|VCD|VCP|VNV|VPT|VRD|VSB)")) {
              hasLexV = true;
            }
          }
          if (hasCC || (hasPU && ! hasLexV)) {
            category += "-CRD";
            //System.out.println("Found coordinate VP"); // testing
          } else if (hasLexV) {
            category += "-COMP";
            //System.out.println("Found complementing VP"); // testing
          } else {
            category += "-ADJT";
            //System.out.println("Found adjoining VP"); // testing
          }
        } else if (chineseSplitVP >= 1) {
          boolean hasBA = false;
          for (Tree kid : kids) {
            if (kid.label().value().startsWith("BA")) {
              hasBA = true;
            } else if (chineseSplitVP == 2 && tlp.basicCategory(kid.label().value()).equals("VP")) {
              for (Tree kidkid : kid.children()) {
                if (kidkid.label().value().startsWith("BA")) {
                  hasBA = true;
                }
              }
            }
          }
          if (hasBA) {
            category += "-BA";
          }
        }
      }

      if (markVPadjunct && baseParentStr.equals("VP")) {
        // cdm 2008: This used to use startsWith("VP") but changed to baseCat
        Tree[] sisters = parent.children();
        boolean hasVPsister = false;
        boolean hasCC = false;
        boolean hasPU = false;
        boolean hasLexV = false;
        for (Tree sister : sisters) {
          if (tlp.basicCategory(sister.label().value()).equals("VP")) {
            hasVPsister = true;
          }
          if (sister.label().value().startsWith("CC")) {
            hasCC = true;
          }
          if (sister.label().value().startsWith("PU")) {
            hasPU = true;
          }
          if (StringUtils.lookingAt(sister.label().value(), "(V[ACEV]|VCD|VCP|VNV|VPT|VRD|VSB)")) {
            hasLexV = true;
          }
        }
        if (hasVPsister && !(hasCC || hasPU || hasLexV)) {
          category += "-VPADJ";
          //System.out.println("Found adjunct of VP"); // testing
        }
      }

      if (markNPmodNP && baseCategory.equals("NP") && baseParentStr.equals("NP")) {
        if (rightSis.contains("NP")) {
          category += "=MODIFIERNP";
          //System.out.println("Found NP modifier of NP"); // testing
        }
      }

      if (markModifiedNP && baseCategory.equals("NP") && baseParentStr.equals("NP")) {
        if (rightSis.isEmpty() && (leftSis.contains("ADJP") || leftSis.contains("NP") || leftSis.contains("DNP") || leftSis.contains("QP") || leftSis.contains("CP") || leftSis.contains("PP"))) {
          category += "=MODIFIEDNP";
          //System.out.println("Found modified NP"); // testing
        }
      }

      if (markNPconj && baseCategory.equals("NP") && baseParentStr.equals("NP")) {
        if (rightSis.contains("CC") || rightSis.contains("PU") || leftSis.contains("CC") || leftSis.contains("PU")) {
          category += "=CONJ";
          //System.out.println("Found NP conjunct"); // testing
        }
      }

      if (markIPconj && baseCategory.equals("IP") && baseParentStr.equals("IP")) {
        Tree[] sisters = parent.children();
        boolean hasCommaSis = false;
        boolean hasIPSis = false;
        for (Tree sister : sisters) {
          if (ctlp.basicCategory(sister.label().value()).equals("PU") && ChineseTreebankLanguagePack.chineseCommaAcceptFilter().test(sister.children()[0].label().toString())) {
            hasCommaSis = true;
            //System.out.println("Found CommaSis"); // testing
          }
          if (ctlp.basicCategory(sister.label().value()).equals("IP") && sister != t) {
            hasIPSis = true;
          }
        }
        if (hasCommaSis && hasIPSis) {
          category += "-CONJ";
          //System.out.println("Found IP conjunct"); // testing
        }
      }

      if (unaryIP && baseCategory.equals("IP") && t.numChildren() == 1) {
        category += "-U";
        //System.out.println("Found unary IP"); //testing
      }
      if (unaryCP && baseCategory.equals("CP") && t.numChildren() == 1) {
        category += "-U";
        //System.out.println("Found unary CP"); //testing
      }

      if (splitBaseNP && baseCategory.equals("NP")) {
        if (t.isPrePreTerminal()) {
          category = category + "-B";
        }
      }

      //if (Test.verbose) printlnErr(baseCategory + " " + leftSis.toString()); //debugging

      if (markPostverbalPP && leftSis.contains("VV") && baseCategory.equals("PP")) {
        //System.out.println("Found post-verbal PP");
        category += "=lVV";
      }

      if ((markADgrandchildOfIP || gpaAD) && listBasicCategories(SisterAnnotationStats.kidLabels(t)).contains("AD")) {
        category += "^ADVP";
      }

      if (markCC) {
        // was: for (int i = 0; i < kids.length; i++) {
        // This second version takes an idea from Collins: don't count
        // marginal conjunctions which don't conjoin 2 things.
        for (int i = 1; i < kids.length - 1; i++) {
          String cat2 = kids[i].label().value();
          if (cat2.startsWith("CC")) {
            category += "-CC";
          }
        }
      }

      Label label = new CategoryWordTag(category, word, tag);
      t.setLabel(label);
    }
    return t;
  }


  /**
   * Chinese: Split the dou hao (a punctuation mark separating
   * members of a list) from other punctuation.  Good but included below.
   */
  public boolean chineseSplitDouHao = false;
  /**
   * Chinese: split Chinese punctuation several ways, along the lines
   * of English punctuation plus another category for the dou hao.  Good.
   */
  public boolean chineseSplitPunct = true;
  /**
   * Chinese: split left right/paren quote (if chineseSplitPunct is also
   * true.  Only very marginal gains, but seems positive.
   */
  public boolean chineseSplitPunctLR = false;

  /**
   * Chinese: mark VVs that are sister of IP (communication &
   * small-clause-taking verbs).  Good: give 0.5%
   */
  public boolean markVVsisterIP = true;

  /**
   * Chinese: mark P's that are sister of IP.  Negative effect
   */
  public boolean markPsisterIP = true;

  /**
   * Chinese: mark IP's that are sister of VV or P.  These rarely
   * have punctuation. Small positive effect.
   */
  public boolean markIPsisterVVorP = true;


  /**
   * Chinese: mark ADs that are grandchild of IP.
   */
  public boolean markADgrandchildOfIP = false;
  /**
   * Grandparent annotate all AD.  Seems slightly negative.
   */
  public boolean gpaAD = true;

  // using tagPA on Chinese 100k is negative.

  public boolean chineseVerySelectiveTagPA = false;
  public boolean chineseSelectiveTagPA = false;

  /**
   * Chinese: mark IPs that are sister of BA.  These always have
   * overt NP.  Very slightly positive.
   */
  public boolean markIPsisterBA = true;

  /**
   * Chinese: mark phrases that are adjuncts of VP (these tend to be
   * locatives/temporals, and have a specific distribution).
   * Necessary even with chineseSplitVP==3 and parent annotation because
   * parent annotation happens with unsplit parent categories.
   * Slightly positive.
   */
  public boolean markVPadjunct = true;

  /**
   * Chinese: mark NP modifiers of NPs. Quite positive (0.5%)
   */
  public boolean markNPmodNP = true;

  /**
   * Chinese: mark left-modified NPs (rightmost NPs with a left-side
   * mod).  Slightly positive.
   */
  public boolean markModifiedNP = true;

  /**
   * Chinese: mark NPs that are conjuncts.  Negative on small set.
   */
  public boolean markNPconj = true;

  /**
   * Chinese: mark nominal tags that are part of multi-nominal
   * rewrites.  Doesn't seem any good.
   */
  public boolean markMultiNtag = false;

  /**
   * Chinese: mark IPs that are part of prenominal modifiers. Negative.
   */
  public boolean markIPsisDEC = true;

  /**
   * Chinese: mark IPs that are conjuncts.  Or those that have
   * (adjuncts or subjects)
   */
  public boolean markIPconj = false;
  public boolean markIPadjsubj = false;

  /**
   * Chinese VP splitting.  0 = none;
   * 1 = mark with -BA a VP that directly dominates a BA;
   * 2 = mark with -BA a VP that directly dominates a BA or a VP that
   *     directly dominates a BA
   * 3 = split VPs into VP-COMP, VP-CRD, VP-ADJ.  (Negative value.)
   */
  public int chineseSplitVP = 3;

  /** Chinese: if an IP has no subject (including no empty-category
   * subject), then it should only have an NP (adjunct) daughter if
   * it's a coordinate IP and the NP scopes over the conjunct
   * IPs. (sometimes this NP daughter is adjoined in an IP -> NP
   * IP_coord structure, sometimes the IP conjuncts are at the same
   * level as the NP).  In other cases NP adjuncts should be inside
   * VP.  So: an IP dominating neither a non-subject NP nor another IP
   * should have no NP daughters.  BUT this generalization breaks down
   * when you try to extend it to IPs ignoring their empty subjects.
   * So the simplest thing to do would be to mark non-subject dtrs of
   * IP....  but I think we need to leave the SBJ functional tagging
   * on categories to be consistent about this.
   *
   * Update: I tried retaining SBJ markers with
   * SbjRetainingTreeNormalizer but it works worse than using
   * markVPadjunct.
   */

  /**
   * Chinese: merge NN and VV.  A lark.
   */
  public boolean mergeNNVV = false;

  // XXXX upto in testing

  /**
   * Chinese: unary category marking
   */
  public boolean unaryIP = false;
  public boolean unaryCP = false;

  /**
   * Chinese: parent annotate daughter of root.  Meant only for
   * selectivesplit=false.
   */
  public boolean paRootDtr = false; // true

  /**
   * Chinese: mark P with a left aunt VV, and PP with a left sister
   * VV.  Note that it's necessary to mark both to thread the
   * context-marking.  Used to identify post-verbal P's, which are
   * rare.
   */
  public boolean markPostverbalP = false;
  public boolean markPostverbalPP = false;


  // Not used now
  // /** How selectively to split. */
  // public int selectiveSplitLevel = 1;

  /**
   * Mark base NPs.  Good.
   */
  public boolean splitBaseNP = false;

  /**
   * Annotate tags for number of characters contained.
   */
  public boolean tagWordSize = false;

  /**
   * Mark phrases which are conjunctions.
   * Appears negative, even with 200K words training data.
   */
  public boolean markCC = false;

  /**
   * Whether to retain the -TMP functional tag on various phrasal
   * categories.  On 80K words training, minutely helpful; on 200K
   * words, best option gives 0.6%.  Doing
   * splitNPTMP and splitPPTMP (but not splitXPTMP) is best.
   */
  public boolean splitNPTMP = false;
  public boolean splitPPTMP = false;
  public boolean splitXPTMP = false;

  /**
   * Verbal distance -- mark whether symbol dominates a verb (V*).
   * Seems bad for Chinese.
   */
  public boolean dominatesV = false;


  /**
   * Parameters specific for creating a ChineseLexicon
   */
  public static final boolean DEFAULT_USE_GOOD_TURNING_UNKNOWN_WORD_MODEL = false;
  public boolean useGoodTuringUnknownWordModel = DEFAULT_USE_GOOD_TURNING_UNKNOWN_WORD_MODEL;
  public boolean useCharBasedUnknownWordModel = false;


  /**
   * Parameters for a ChineseCharacterBasedLexicon
   */
  public double lengthPenalty = 5.0;

  public boolean useUnknownCharacterModel = true;

  /**
   * penaltyType should be set as follows:
   * 0: no length penalty
   * 1: quadratic length penalty
   * 2: penalty for continuation chars only
   * TODO: make this an enum
   */
  public int penaltyType = 0;

  @Override
  public void display() {
    String chineseParams = "Using ChineseTreebankParserParams" + " chineseSplitDouHao=" + chineseSplitDouHao + " chineseSplitPunct=" + chineseSplitPunct + " chineseSplitPunctLR=" + chineseSplitPunctLR + " markVVsisterIP=" + markVVsisterIP + " markVPadjunct=" + markVPadjunct + " chineseSplitVP=" + chineseSplitVP + " mergeNNVV=" + mergeNNVV + " unaryIP=" + unaryIP + " unaryCP=" + unaryCP + " paRootDtr=" + paRootDtr + " markPsisterIP=" + markPsisterIP + " markIPsisterVVorP=" + markIPsisterVVorP + " markADgrandchildOfIP=" + markADgrandchildOfIP + " gpaAD=" + gpaAD + " markIPsisterBA=" + markIPsisterBA + " markNPmodNP=" + markNPmodNP + " markNPconj=" + markNPconj + " markMultiNtag=" + markMultiNtag + " markIPsisDEC=" + markIPsisDEC + " markIPconj=" + markIPconj + " markIPadjsubj=" + markIPadjsubj + " markPostverbalP=" + markPostverbalP + " markPostverbalPP=" + markPostverbalPP
            //      + " selSplitLevel=" + selectiveSplitLevel
            + " baseNP=" + splitBaseNP + " headFinder=" + (sunJurafskyHeadFinder ? "sunJurafsky" : (bikelHeadFinder ? "bikel" : "levy")) + " discardFrags=" + discardFrags  + " dominatesV=" + dominatesV;
    printlnErr(chineseParams);
  }


  private List<String> listBasicCategories(List<String> l) {
    List<String> l1 = new ArrayList<String>();
    for (String s : l) {
      l1.add(ctlp.basicCategory(s));
    }
    return l1;
  }

  // TODO: Rewrite this as general matching predicate
  private static boolean hasV(List tags) {
    for (int i = 0, tsize = tags.size(); i < tsize; i++) {
      String str = tags.get(i).toString();
      if (str.startsWith("V")) {
        return true;
      }
    }
    return false;
  }

  /**
   * Set language-specific options according to flags.
   * This routine should process the option starting in args[i] (which
   * might potentially be several arguments long if it takes arguments).
   * It should return the index after the last index it consumed in
   * processing.  In particular, if it cannot process the current option,
   * the return value should be i.
   */
  @Override
  public int setOptionFlag(String[] args, int i) {
    // [CDM 2008: there are no generic options!] first, see if it's a generic option
    // int j = super.setOptionFlag(args, i);
    // if(i != j) return j;

    //lang. specific options
    // if (args[i].equalsIgnoreCase("-vSelSplitLevel") &&
    //            (i+1 < args.length)) {
    //   selectiveSplitLevel = Integer.parseInt(args[i+1]);
    //   i+=2;
    // } else
    if (args[i].equalsIgnoreCase("-paRootDtr")) {
      paRootDtr = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-unaryIP")) {
      unaryIP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-unaryCP")) {
      unaryCP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markPostverbalP")) {
      markPostverbalP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markPostverbalPP")) {
      markPostverbalPP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-baseNP")) {
      splitBaseNP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markVVsisterIP")) {
      markVVsisterIP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markPsisterIP")) {
      markPsisterIP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markIPsisterVVorP")) {
      markIPsisterVVorP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markIPsisterBA")) {
      markIPsisterBA = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-dominatesV")) {
      dominatesV = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-gpaAD")) {
      gpaAD = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markVPadjunct")) {
      markVPadjunct = Boolean.valueOf(args[i+1]);
      i += 2;
    } else if (args[i].equalsIgnoreCase("-markNPmodNP")) {
      markNPmodNP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markModifiedNP")) {
      markModifiedNP = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-nomarkModifiedNP")) {
      markModifiedNP = false;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markNPconj")) {
      markNPconj = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-nomarkNPconj")) {
      markNPconj = false;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-chineseSplitPunct")) {
      chineseSplitPunct = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-chineseSplitPunctLR")) {
      chineseSplitPunct = true;
      chineseSplitPunctLR = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-chineseSelectiveTagPA")) {
      chineseSelectiveTagPA = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-chineseVerySelectiveTagPA")) {
      chineseVerySelectiveTagPA = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-markIPsisDEC")) {
      markIPsisDEC = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-chineseSplitVP")) {
      chineseSplitVP = Integer.parseInt(args[i+1]);
      i += 2;
    } else if (args[i].equalsIgnoreCase("-tagWordSize")) {
      tagWordSize = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-vanilla")) {
      chineseSplitDouHao = false;
      chineseSplitPunct = false;
      chineseSplitPunctLR = false;
      markVVsisterIP = false;
      markPsisterIP = false;
      markIPsisterVVorP = false;
      markADgrandchildOfIP = false;
      gpaAD = false;
      markIPsisterBA = false;
      markVPadjunct = false;
      markNPmodNP = false;
      markModifiedNP = false;
      markNPconj = false;
      markMultiNtag = false;
      markIPsisDEC = false;
      markIPconj = false;
      markIPadjsubj = false;
      chineseSplitVP = 0;
      mergeNNVV = false;
      unaryIP = false;
      unaryCP = false;
      paRootDtr = false;
      markPostverbalP = false;
      markPostverbalPP = false;
      splitBaseNP = false;
      // selectiveSplitLevel = 0;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-acl03chinese")) {
      chineseSplitDouHao = false;
      chineseSplitPunct = true;
      chineseSplitPunctLR = true;
      markVVsisterIP = true;
      markPsisterIP = true;
      markIPsisterVVorP = true;
      markADgrandchildOfIP = false;
      gpaAD = true;
      markIPsisterBA = false;
      markVPadjunct = true;
      markNPmodNP = true;
      markModifiedNP = true;
      markNPconj = true;
      markMultiNtag = false;
      markIPsisDEC = true;
      markIPconj = false;
      markIPadjsubj = false;
      chineseSplitVP = 3;
      mergeNNVV = false;
      unaryIP = true;
      unaryCP = true;
      paRootDtr = true;
      markPostverbalP = false;
      markPostverbalPP = false;
      splitBaseNP = false;
      // selectiveSplitLevel = 0;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-chineseFactored")) {
      chineseSplitDouHao = false;
      chineseSplitPunct = true;
      chineseSplitPunctLR = true;
      markVVsisterIP = true;
      markPsisterIP = true;
      markIPsisterVVorP = true;
      markADgrandchildOfIP = false;
      gpaAD = true;
      markIPsisterBA = true;
      markVPadjunct = true;
      markNPmodNP = true;
      markModifiedNP = true;
      markNPconj = true;
      markMultiNtag = false;
      markIPsisDEC = true;
      markIPconj = false;
      markIPadjsubj = false;
      chineseSplitVP = 3;
      mergeNNVV = false;
      unaryIP = true;
      unaryCP = true;
      paRootDtr = true;
      markPostverbalP = false;
      markPostverbalPP = false;
      splitBaseNP = false;
      // selectiveSplitLevel = 0;
      chineseVerySelectiveTagPA = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-chinesePCFG")) {
      chineseSplitDouHao = false;
      chineseSplitPunct = true;
      chineseSplitPunctLR = true;
      markVVsisterIP = true;
      markPsisterIP = false;
      markIPsisterVVorP = true;
      markADgrandchildOfIP = false;
      gpaAD = false;
      markIPsisterBA = true;
      markVPadjunct = true;
      markNPmodNP = true;
      markModifiedNP = true;
      markNPconj = false;
      markMultiNtag = false;
      markIPsisDEC = false;
      markIPconj = false;
      markIPadjsubj = false;
      chineseSplitVP = 0;
      mergeNNVV = false;
      unaryIP = false;
      unaryCP = false;
      paRootDtr = false;
      markPostverbalP = false;
      markPostverbalPP = false;
      splitBaseNP = false;
      // selectiveSplitLevel = 0;
      chineseVerySelectiveTagPA = true;
      i += 1;
    } else if (args[i].equalsIgnoreCase("-sunHead")) {
      sunJurafskyHeadFinder = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-bikelHead")) {
      bikelHeadFinder = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-discardFrags")) {
      discardFrags = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-charLex")) {
      useCharacterBasedLexicon = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-charUnk")) {
      useCharBasedUnknownWordModel = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-rad")) {
      useUnknownCharacterModel = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-lengthPenalty") && (i + 1 < args.length)) {
      lengthPenalty = Double.parseDouble(args[i + 1]);
      i += 2;
    } else if (args[i].equalsIgnoreCase("-penaltyType") && (i + 1 < args.length)) {
      penaltyType = Integer.parseInt(args[i + 1]);
      i += 2;
    } else if (args[i].equalsIgnoreCase("-gtUnknown")) {
      useGoodTuringUnknownWordModel = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-maxentUnk")) {
      // useMaxentUnknownWordModel = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-tuneSigma")) {
      // ChineseMaxentLexicon.tuneSigma = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-trainCountThresh") && (i + 1 < args.length)) {
      // ChineseMaxentLexicon.trainCountThreshold = Integer.parseInt(args[i + 1]);
      i += 2;
    } else if (args[i].equalsIgnoreCase("-markCC")) {
      markCC = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-segmentMarkov") || args[i].equalsIgnoreCase("-segmentWords")) {
      segment = true;
      segmentMarkov = true;
      segmenterClass = "edu.stanford.nlp.parser.lexparser.ChineseMarkovWordSegmenter";
      i++;
    } else if (args[i].equalsIgnoreCase("-segmentMaxMatch")) {
      segment = true;
      segmentMarkov = false;
      segmenterClass = "edu.stanford.nlp.parser.lexparser.MaxMatchSegmenter";
      i++;
    } else if (args[i].equalsIgnoreCase("-segmentDPMaxMatch")) {
      segment = true;
      segmentMarkov = false;
      segmenterClass = "edu.stanford.nlp.wordseg.MaxMatchSegmenter";
      i++;
    } else if (args[i].equalsIgnoreCase("-maxentLex")) {
      // useMaxentLexicon = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-fixUnkFunctionWords")) {
      // ChineseMaxentLexicon.fixUnkFunctionWords = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-similarWordSmoothing")) {
      useSimilarWordMap = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-maxentLexSeenTagsOnly")) {
      // useMaxentLexicon = true;
      // ChineseMaxentLexicon.seenTagsOnly = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-maxentLexFeatLevel") && (i + 1 < args.length)) {
      // ChineseMaxentLexicon.featureLevel = Integer.parseInt(args[i + 1]);
      i += 2;
    } else if (args[i].equalsIgnoreCase("-maxentDepGrammarFeatLevel") && (i + 1 < args.length)) {
      depGramFeatureLevel = Integer.parseInt(args[i + 1]);
      i += 2;
    } else if (args[i].equalsIgnoreCase("-maxentDepGrammar")) {
      // useMaxentDepGrammar = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-splitNPTMP")) {
      splitNPTMP = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-splitPPTMP")) {
      splitPPTMP = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-splitXPTMP")) {
      splitXPTMP = true;
      i++;
    } else if (args[i].equalsIgnoreCase("-segmenter")) {
      segment = true;
      segmentMarkov = false;
      segmenterClass = args[i + 1];
      i += 2;
    } else if (args[i].equalsIgnoreCase("-headFinder") && (i + 1 < args.length)) {
      try {
        headFinder = (HeadFinder) Class.forName(args[i + 1]).newInstance();
      } catch (Exception e) {
        System.err.println(e);
        System.err.println(this.getClass().getName() + ": Could not load head finder " + args[i + 1]);
        throw new RuntimeException(e);
      }
      i+=2;
    }

    return i;
  }

  private int depGramFeatureLevel = 0;

  @Override
  public Extractor<DependencyGrammar> dependencyGrammarExtractor(final Options op, Index<String> wordIndex, Index<String> tagIndex) {
    /* ----------
    if (useMaxentDepGrammar) {
      return new Extractor() {
        public Object extract(Collection<Tree> trees) {
          ChineseWordFeatureExtractor wfe = new ChineseWordFeatureExtractor(trees);
          ChineseWordFeatureExtractor wfe2 = new ChineseWordFeatureExtractor(trees);
          wfe.setFeatureLevel(2);
          wfe2.turnOffWordFeatures = true;
          wfe2.setFeatureLevel(depGramFeatureLevel);
          MaxentDependencyGrammar dg = new MaxentDependencyGrammar(op.tlpParams, wfe, wfe2, true, false, false);
          dg.train(trees);
          return dg;
        }

        public Object extract(Iterator<Tree> iterator, Function<Tree, Tree> f) {
          throw new UnsupportedOperationException();
        }
      };
    } else ------- */
    if (useSimilarWordMap) {
      return new MLEDependencyGrammarExtractor(op, wordIndex, tagIndex) {
        @Override
        public MLEDependencyGrammar formResult() {
          wordIndex.addToIndex(Lexicon.UNKNOWN_WORD);
          ChineseSimWordAvgDepGrammar dg = new ChineseSimWordAvgDepGrammar(tlpParams, directional, useDistance, useCoarseDistance, op.trainOptions.basicCategoryTagsInDependencyGrammar, op, wordIndex, tagIndex);
          if (lex == null) {
            throw new RuntimeException("Attempt to create ChineseSimWordAvgDepGrammar before Lexicon!!!");
          } else {
            dg.setLex(lex);
          }
          for (IntDependency dependency : dependencyCounter.keySet()) {
            dg.addRule(dependency, dependencyCounter.getCount(dependency));
          }
          return dg;
        }

     };
    } else {
      return new MLEDependencyGrammarExtractor(op, wordIndex, tagIndex);
    }
  }

  /**
   * Return a default sentence for the language (for testing)
   */
  @Override
  public ArrayList<Word> defaultTestSentence() {
    return Sentence.toUntaggedList("\u951f\u65a4\u62f7", "\u951f\u65a4\u62f7", "\u5b66\u6821", "\u951f\u65a4\u62f7", "\u5b66\u4e60", "\u951f\u65a4\u62f7");
  }


  private static final long serialVersionUID = 2;


  @Override
  public List<GrammaticalStructure>
    readGrammaticalStructureFromFile(String filename)
  {
    try {
      return ChineseGrammaticalStructure.
              readCoNLLXGrammaticalStructureCollection(filename);
    } catch (IOException e) {
      throw new RuntimeIOException(e);
    }
  }

  @Override
  public GrammaticalStructure getGrammaticalStructure(Tree t,
                                                      Predicate<String> filter,
                                                      HeadFinder hf) {
    return new ChineseGrammaticalStructure(t, filter, hf);
  }

  @Override
  public boolean supportsBasicDependencies() {
    return true;
  }

  /**
   * For testing: loads a treebank and prints the trees.
   */
  public static void main(String[] args) {
    TreebankLangParserParams tlpp = new ChineseTreebankParserParams();
    System.out.println("Default encoding is: " +
                       tlpp.diskTreebank().encoding());

    if (args.length < 2) {
      printlnErr("Usage: edu.stanford.nlp.parser.lexparser.ChineseTreebankParserParams treesPath fileRange");
    } else {
      Treebank m = tlpp.diskTreebank();
      m.loadPath(args[0], new NumberRangesFileFilter(args[1], false));

      for (Tree t : m ) {
        t.pennPrint(tlpp.pw());
      }
      System.out.println("There were " + m.size() + " trees.");
    }
  }

}
TOP

Related Classes of edu.stanford.nlp.parser.lexparser.ChineseTreebankParserParams

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.