LUCENE-5311 - added support for training using multiple content fields for knn and naive bayes

git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/trunk@1538205 13f79535-47bb-0310-9956-ffa450edef68

lucene/classification/src/java/org/apache/lucene/classification/BooleanPerceptronClassifier.java

@@ -191,6 +191,11 @@ public class BooleanPerceptronClassifier implements Classifier<Boolean> {
     weights.clear(); // free memory while waiting for GC
   }
 
+  @Override
+  public void train(AtomicReader atomicReader, String[] textFieldNames, String classFieldName, Analyzer analyzer, Query query) throws IOException {
+    throw new IOException("training with multiple fields not supported by boolean perceptron classifier");
+  }
+
   private TermsEnum updateWeights(AtomicReader atomicReader, TermsEnum reuse,
       int docId, Boolean assignedClass, SortedMap<String,Double> weights,
       double modifier, boolean updateFST) throws IOException {

lucene/classification/src/java/org/apache/lucene/classification/ClassificationResult.java

@@ -22,8 +22,8 @@ package org.apache.lucene.classification;
  */
 public class ClassificationResult<T> {
 
-  private T assignedClass;
+  private final T assignedClass;
-  private double score;
+  private final double score;
 
   /**
    * Constructor

lucene/classification/src/java/org/apache/lucene/classification/Classifier.java

@@ -60,4 +60,16 @@ public interface Classifier<T> {
   public void train(AtomicReader atomicReader, String textFieldName, String classFieldName, Analyzer analyzer, Query query)
       throws IOException;
 
+  /**
+   * Train the classifier using the underlying Lucene index
+   * @param atomicReader the reader to use to access the Lucene index
+   * @param textFieldNames the names of the fields to be used to compare documents
+   * @param classFieldName the name of the field containing the class assigned to documents
+   * @param analyzer the analyzer used to tokenize / filter the unseen text
+   * @param query the query to filter which documents use for training
+   * @throws IOException If there is a low-level I/O error.
+   */
+  public void train(AtomicReader atomicReader, String[] textFieldNames, String classFieldName, Analyzer analyzer, Query query)
+      throws IOException;
+
 }

lucene/classification/src/java/org/apache/lucene/classification/KNearestNeighborClassifier.java

@@ -41,10 +41,10 @@ import java.util.Map;
 public class KNearestNeighborClassifier implements Classifier<BytesRef> {
 
   private MoreLikeThis mlt;
-  private String textFieldName;
+  private String[] textFieldNames;
   private String classFieldName;
   private IndexSearcher indexSearcher;
-  private int k;
+  private final int k;
   private Query query;
 
   /**
@@ -65,14 +65,17 @@ public class KNearestNeighborClassifier implements Classifier<BytesRef> {
       throw new IOException("You must first call Classifier#train");
     }
     Query q;
+    BooleanQuery mltQuery = new BooleanQuery();
+    for (String textFieldName : textFieldNames) {
+      mltQuery.add(new BooleanClause(mlt.like(new StringReader(text), textFieldName), BooleanClause.Occur.SHOULD));
+    }
     if (query != null) {
-      Query mltQuery = mlt.like(new StringReader(text), textFieldName);
       BooleanQuery bq = new BooleanQuery();
       bq.add(query, BooleanClause.Occur.MUST);
       bq.add(mltQuery, BooleanClause.Occur.MUST);
       q = bq;
     } else {
-      q = mlt.like(new StringReader(text), textFieldName);
+      q = mltQuery;
     }
     TopDocs topDocs = indexSearcher.search(q, k);
     return selectClassFromNeighbors(topDocs);
@@ -116,7 +119,7 @@ public class KNearestNeighborClassifier implements Classifier<BytesRef> {
    */
   @Override
   public void train(AtomicReader atomicReader, String textFieldName, String classFieldName, Analyzer analyzer, Query query) throws IOException {
-    this.textFieldName = textFieldName;
+    this.textFieldNames = new String[]{textFieldName};
     this.classFieldName = classFieldName;
     mlt = new MoreLikeThis(atomicReader);
     mlt.setAnalyzer(analyzer);
@@ -124,4 +127,18 @@ public class KNearestNeighborClassifier implements Classifier<BytesRef> {
     indexSearcher = new IndexSearcher(atomicReader);
     this.query = query;
   }
+
+  /**
+   * {@inheritDoc}
+   */
+  @Override
+  public void train(AtomicReader atomicReader, String[] textFieldNames, String classFieldName, Analyzer analyzer, Query query) throws IOException {
+    this.textFieldNames = textFieldNames;
+    this.classFieldName = classFieldName;
+    mlt = new MoreLikeThis(atomicReader);
+    mlt.setAnalyzer(analyzer);
+    mlt.setFieldNames(textFieldNames);
+    indexSearcher = new IndexSearcher(atomicReader);
+    this.query = query;
+  }
 }

lucene/classification/src/java/org/apache/lucene/classification/SimpleNaiveBayesClassifier.java

@@ -45,7 +45,7 @@ import java.util.LinkedList;
 public class SimpleNaiveBayesClassifier implements Classifier<BytesRef> {
 
   private AtomicReader atomicReader;
-  private String textFieldName;
+  private String[] textFieldNames;
   private String classFieldName;
   private int docsWithClassSize;
   private Analyzer analyzer;
@@ -68,18 +68,36 @@ public class SimpleNaiveBayesClassifier implements Classifier<BytesRef> {
       throws IOException {
     this.atomicReader = atomicReader;
     this.indexSearcher = new IndexSearcher(this.atomicReader);
-    this.textFieldName = textFieldName;
+    this.textFieldNames = new String[]{textFieldName};
     this.classFieldName = classFieldName;
     this.analyzer = analyzer;
     this.docsWithClassSize = countDocsWithClass();
     this.query = query;
   }
 
+  /**
+   * {@inheritDoc}
+   */
   @Override
   public void train(AtomicReader atomicReader, String textFieldName, String classFieldName, Analyzer analyzer) throws IOException {
     train(atomicReader, textFieldName, classFieldName, analyzer, null);
   }
 
+  /**
+   * {@inheritDoc}
+   */
+  @Override
+  public void train(AtomicReader atomicReader, String[] textFieldNames, String classFieldName, Analyzer analyzer, Query query)
+      throws IOException {
+    this.atomicReader = atomicReader;
+    this.indexSearcher = new IndexSearcher(this.atomicReader);
+    this.textFieldNames = textFieldNames;
+    this.classFieldName = classFieldName;
+    this.analyzer = analyzer;
+    this.docsWithClassSize = countDocsWithClass();
+    this.query = query;
+  }
+
   private int countDocsWithClass() throws IOException {
     int docCount = MultiFields.getTerms(this.atomicReader, this.classFieldName).getDocCount();
     if (docCount == -1) { // in case codec doesn't support getDocCount
@@ -103,6 +121,7 @@ public class SimpleNaiveBayesClassifier implements Classifier<BytesRef> {
 
   private String[] tokenizeDoc(String doc) throws IOException {
     Collection<String> result = new LinkedList<String>();
+    for (String textFieldName : textFieldNames) {
       try (TokenStream tokenStream = analyzer.tokenStream(textFieldName, doc)) {
         CharTermAttribute charTermAttribute = tokenStream.addAttribute(CharTermAttribute.class);
         tokenStream.reset();
@@ -111,6 +130,7 @@ public class SimpleNaiveBayesClassifier implements Classifier<BytesRef> {
         }
         tokenStream.end();
       }
+    }
     return result.toArray(new String[result.size()]);
   }
 
@@ -164,16 +184,23 @@ public class SimpleNaiveBayesClassifier implements Classifier<BytesRef> {
   }
 
   private double getTextTermFreqForClass(BytesRef c) throws IOException {
+    double avgNumberOfUniqueTerms = 0;
+    for (String textFieldName : textFieldNames) {
       Terms terms = MultiFields.getTerms(atomicReader, textFieldName);
       long numPostings = terms.getSumDocFreq(); // number of term/doc pairs
-    double avgNumberOfUniqueTerms = numPostings / (double) terms.getDocCount(); // avg # of unique terms per doc
+      avgNumberOfUniqueTerms += numPostings / (double) terms.getDocCount(); // avg # of unique terms per doc
+    }
     int docsWithC = atomicReader.docFreq(new Term(classFieldName, c));
     return avgNumberOfUniqueTerms * docsWithC; // avg # of unique terms in text field per doc * # docs with c
   }
 
   private int getWordFreqForClass(String word, BytesRef c) throws IOException {
     BooleanQuery booleanQuery = new BooleanQuery();
-    booleanQuery.add(new BooleanClause(new TermQuery(new Term(textFieldName, word)), BooleanClause.Occur.MUST));
+    BooleanQuery subQuery = new BooleanQuery();
+    for (String textFieldName : textFieldNames) {
+     subQuery.add(new BooleanClause(new TermQuery(new Term(textFieldName, word)), BooleanClause.Occur.SHOULD));
+    }
+    booleanQuery.add(new BooleanClause(subQuery, BooleanClause.Occur.MUST));
     booleanQuery.add(new BooleanClause(new TermQuery(new Term(classFieldName, c)), BooleanClause.Occur.MUST));
     if (query != null) {
       booleanQuery.add(query, BooleanClause.Occur.MUST);

lucene/classification/src/java/org/apache/lucene/classification/utils/DatasetSplitter.java

@@ -40,8 +40,8 @@ import java.io.IOException;
  */
 public class DatasetSplitter {
 
-  private double crossValidationRatio;
+  private final double crossValidationRatio;
-  private double testRatio;
+  private final double testRatio;
 
   /**
    * Create a {@link DatasetSplitter} by giving test and cross validation IDXs sizes
@@ -68,8 +68,6 @@ public class DatasetSplitter {
   public void split(AtomicReader originalIndex, Directory trainingIndex, Directory testIndex, Directory crossValidationIndex,
                     Analyzer analyzer, String... fieldNames) throws IOException {
 
-    // TODO : check that the passed fields are stored in the original index
-
     // create IWs for train / test / cv IDXs
     IndexWriter testWriter = new IndexWriter(testIndex, new IndexWriterConfig(Version.LUCENE_50, analyzer));
     IndexWriter cvWriter = new IndexWriter(crossValidationIndex, new IndexWriterConfig(Version.LUCENE_50, analyzer));