edu.stanford.nlp.ling.Word



API Populatity

1 Client projects

Project: edu.stanford

Package: edu.stanford.nlp

Project stanfordnlp/CoreNLP in file ...src.edu.stanford.nlp.pipeline.MorphaAnnotator.java (2014-05-19)
@@ -3,7 +3,6 @@ package edu.stanford.nlp.pipeline;
 import edu.stanford.nlp.ling.CoreAnnotation;
 import edu.stanford.nlp.ling.CoreAnnotations;
 import edu.stanford.nlp.ling.CoreLabel;
-import edu.stanford.nlp.ling.Word;
 import edu.stanford.nlp.process.Morphology;
 import edu.stanford.nlp.util.CoreMap;
 
@@ -53,7 +52,7 @@ public class MorphaAnnotator implements Annotator{
         for (CoreLabel token : tokens) {
           String text = token.get(CoreAnnotations.TextAnnotation.class);
           String posTag = token.get(CoreAnnotations.PartOfSpeechAnnotation.class);
-          this.addLemma(morphology, CoreAnnotations.LemmaAnnotation.class, token, text, posTag);
+          addLemma(morphology, CoreAnnotations.LemmaAnnotation.class, token, text, posTag);
         }
       }
     } else {
Project stanfordnlp/CoreNLP in file ...src.edu.stanford.nlp.pipeline.MorphaAnnotator.java (2014-05-19)
@@ -3,6 +3,7 @@ package edu.stanford.nlp.pipeline;
 import edu.stanford.nlp.ling.CoreAnnotation;
 import edu.stanford.nlp.ling.CoreAnnotations;
 import edu.stanford.nlp.ling.CoreLabel;
+import edu.stanford.nlp.ling.Word;
 import edu.stanford.nlp.process.Morphology;
 import edu.stanford.nlp.util.CoreMap;
 
@@ -52,7 +53,7 @@ public class MorphaAnnotator implements Annotator{
         for (CoreLabel token : tokens) {
           String text = token.get(CoreAnnotations.TextAnnotation.class);
           String posTag = token.get(CoreAnnotations.PartOfSpeechAnnotation.class);
-          addLemma(morphology, CoreAnnotations.LemmaAnnotation.class, token, text, posTag);
+          this.addLemma(morphology, CoreAnnotations.LemmaAnnotation.class, token, text, posTag);
         }
       }
     } else {
Project stanfordnlp/CoreNLP in file ...src.edu.stanford.nlp.pipeline.MorphaAnnotator.java (2014-02-05)
@@ -3,6 +3,7 @@ package edu.stanford.nlp.pipeline;
 import edu.stanford.nlp.ling.CoreAnnotation;
 import edu.stanford.nlp.ling.CoreAnnotations;
 import edu.stanford.nlp.ling.CoreLabel;
+import edu.stanford.nlp.ling.Word;
 import edu.stanford.nlp.process.Morphology;
 import edu.stanford.nlp.util.CoreMap;
 
@@ -52,7 +53,7 @@ public class MorphaAnnotator implements Annotator{
         for (CoreLabel token : tokens) {
           String text = token.get(CoreAnnotations.TextAnnotation.class);
           String posTag = token.get(CoreAnnotations.PartOfSpeechAnnotation.class);
-          addLemma(morphology, CoreAnnotations.LemmaAnnotation.class, token, text, posTag);
+          this.addLemma(morphology, CoreAnnotations.LemmaAnnotation.class, token, text, posTag);
         }
       }
     } else {