Add an option to skip the MWT in a conllu file when training a tagger

Author: AngledLuffa

src/edu/stanford/nlp/tagger/io/TSVTaggedFileReader.java

@@ -16,6 +16,8 @@ public class TSVTaggedFileReader implements TaggedFileReader {
   private final String filename;
   private final int wordColumn, tagColumn;
   private final boolean usesComments;
+  // if in a conllu file from UD used directly, might want to skip MWT
+  private final boolean skipMWT;
   private List<TaggedWord> next; // = null;
   private int linesRead; // = 0;
 
@@ -36,6 +38,7 @@ public TSVTaggedFileReader(TaggedFileRecord record) {
     tagColumn = ((record.tagColumn == null) ?
                  DEFAULT_TAG_COLUMN : record.tagColumn);
     usesComments = record.usesComments;
+    skipMWT = record.skipMWT;
     primeNext();
   }
 
@@ -85,9 +88,11 @@ private void primeNext() {
           throw new IllegalArgumentException("File " + filename + " line #" +
                                              linesRead + " too short");
         }
-        String word = pieces[wordColumn];
-        String tag = pieces[tagColumn];
-        next.add(new TaggedWord(word, tag));
+        if (!(skipMWT && pieces[0].matches("[0-9]+-[0-9]+"))) {
+          String word = pieces[wordColumn];
+          String tag = pieces[tagColumn];
+          next.add(new TaggedWord(word, tag));
+        }
       }
       try {
         line = reader.readLine();

src/edu/stanford/nlp/tagger/io/TaggedFileRecord.java

@@ -38,6 +38,7 @@ public enum Format {
   final Integer tagColumn;
   final TreeReaderFactory trf;
   final boolean usesComments;
+  final boolean skipMWT;
 
   private TaggedFileRecord(String file, Format format,
                            String encoding, String tagSeparator,
@@ -47,7 +48,7 @@ private TaggedFileRecord(String file, Format format,
                            NumberRangesFileFilter treeRange,
                            Predicate<Tree> treeFilter,
                            Integer wordColumn, Integer tagColumn,
-                           boolean usesComments) {
+                           boolean usesComments, boolean skipMWT) {
     this.file = file;
     this.format = format;
     this.encoding = encoding;
@@ -60,6 +61,7 @@ private TaggedFileRecord(String file, Format format,
     this.tagColumn = tagColumn;
     this.trf = trf;
     this.usesComments = usesComments;
+    this.skipMWT = skipMWT;
   }
 
   public static final String FORMAT = "format";
@@ -73,6 +75,7 @@ private TaggedFileRecord(String file, Format format,
   public static final String TAG_COLUMN = "tagColumn";
   public static final String TREE_READER = "trf";
   public static final String COMMENTS = "comments";
+  public static final String SKIP_MWT = "skipMWT";
 
   public String toString() {
     StringBuilder s = new StringBuilder();
@@ -107,6 +110,9 @@ public String toString() {
     if (usesComments) {
       s.append("," + COMMENTS + "=true");
     }
+    if (skipMWT) {
+      s.append("," + SKIP_MWT + "=true");
+    }
     return s.toString();
   }
 
@@ -142,7 +148,7 @@ public static TaggedFileRecord createRecord(Properties config,
       return new TaggedFileRecord(description, Format.TEXT,
                                   getEncoding(config),
                                   getTagSeparator(config),
-                                  null, null, null, null, null, null, null, false);
+                                  null, null, null, null, null, null, null, false, false);
     }
 
     String[] args = new String[pieces.length - 1];
@@ -158,6 +164,7 @@ public static TaggedFileRecord createRecord(Properties config,
     Predicate<Tree> treeFilter = null;
     Integer wordColumn = null, tagColumn = null;
     boolean comments = false;
+    boolean skipMWT = false;
 
     for (String arg : args) {
       String[] argPieces = arg.split("=", 2);
@@ -188,14 +195,16 @@ public static TaggedFileRecord createRecord(Properties config,
         tagColumn = Integer.valueOf(argPieces[1]);
       } else if (argPieces[0].equalsIgnoreCase(COMMENTS)) {
         comments = Boolean.valueOf(argPieces[1]);
+      } else if (argPieces[0].equalsIgnoreCase(SKIP_MWT)) {
+        skipMWT = Boolean.valueOf(argPieces[1]);
       } else {
         throw new IllegalArgumentException("TaggedFileRecord argument " +
                                            argPieces[0] + " is unknown");
       }
     }
     return new TaggedFileRecord(file, format, encoding, tagSeparator,
                                 treeTransformer, treeNormalizer, trf, treeRange,
-                                treeFilter, wordColumn, tagColumn, comments);
+                                treeFilter, wordColumn, tagColumn, comments, skipMWT);
   }
 
   public static String getEncoding(Properties config) {