[46/50] [abbrv] opennlp git commit: OPENNLP-1010: Fix NaiveBayes model writer

[jo...@apache.org]

OPENNLP-1010: Fix NaiveBayes model writer

The previous sortValues method was based on Perceptron, but for some reason it was not working
for NaiveBayes. Changed it to the one from GIS fixed it.

this closes apache/opennlp#154


Project: http://git-wip-us.apache.org/repos/asf/opennlp/repo
Commit: http://git-wip-us.apache.org/repos/asf/opennlp/commit/15862426
Tree: http://git-wip-us.apache.org/repos/asf/opennlp/tree/15862426
Diff: http://git-wip-us.apache.org/repos/asf/opennlp/diff/15862426

Branch: refs/heads/parser_regression
Commit: 158624265fef4269e2b506ddf708c4b38d2f331d
Parents: 7b1cb70
Author: William D C M SILVA <co...@apache.org>
Authored: Fri Apr 14 09:35:36 2017 -0300
Committer: J�rn Kottmann <jo...@apache.org>
Committed: Thu Apr 20 12:40:25 2017 +0200

----------------------------------------------------------------------
 .../ml/naivebayes/NaiveBayesModelWriter.java    |  71 ++++---
 .../NaiveBayesSerializedCorrectnessTest.java    | 184 +++++++++++++++++++
 2 files changed, 225 insertions(+), 30 deletions(-)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/opennlp/blob/15862426/opennlp-tools/src/main/java/opennlp/tools/ml/naivebayes/NaiveBayesModelWriter.java
----------------------------------------------------------------------
diff --git a/opennlp-tools/src/main/java/opennlp/tools/ml/naivebayes/NaiveBayesModelWriter.java b/opennlp-tools/src/main/java/opennlp/tools/ml/naivebayes/NaiveBayesModelWriter.java
index bbb6eee..510bf76 100644
--- a/opennlp-tools/src/main/java/opennlp/tools/ml/naivebayes/NaiveBayesModelWriter.java
+++ b/opennlp-tools/src/main/java/opennlp/tools/ml/naivebayes/NaiveBayesModelWriter.java
@@ -55,44 +55,55 @@ public abstract class NaiveBayesModelWriter extends AbstractModelWriter {
     }
   }
 
+
   protected ComparablePredicate[] sortValues() {
-    ComparablePredicate[] sortPreds;
-    ComparablePredicate[] tmpPreds = new ComparablePredicate[PARAMS.length];
-    int[] tmpOutcomes = new int[numOutcomes];
-    double[] tmpParams = new double[numOutcomes];
-    int numPreds = 0;
-    //remove parameters with 0 weight and predicates with no parameters
-    for (int pid = 0; pid < PARAMS.length; pid++) {
-      int numParams = 0;
-      double[] predParams = PARAMS[pid].getParameters();
-      int[] outcomePattern = PARAMS[pid].getOutcomes();
-      for (int pi = 0; pi < predParams.length; pi++) {
-        if (predParams[pi] != 0d) {
-          tmpOutcomes[numParams] = outcomePattern[pi];
-          tmpParams[numParams] = predParams[pi];
-          numParams++;
-        }
-      }
 
-      int[] activeOutcomes = new int[numParams];
-      double[] activeParams = new double[numParams];
+    ComparablePredicate[] sortPreds = new ComparablePredicate[PARAMS.length];
 
-      for (int pi = 0; pi < numParams; pi++) {
-        activeOutcomes[pi] = tmpOutcomes[pi];
-        activeParams[pi] = tmpParams[pi];
-      }
-      if (numParams != 0) {
-        tmpPreds[numPreds] = new ComparablePredicate(PRED_LABELS[pid], activeOutcomes, activeParams);
-        numPreds++;
-      }
+    int numParams = 0;
+    for (int pid = 0; pid < PARAMS.length; pid++) {
+      int[] predkeys = PARAMS[pid].getOutcomes();
+      // Arrays.sort(predkeys);
+      int numActive = predkeys.length;
+      double[] activeParams = PARAMS[pid].getParameters();
+
+      numParams += numActive;
+      /*
+       * double[] activeParams = new double[numActive];
+       *
+       * int id = 0; for (int i=0; i < predkeys.length; i++) { int oid =
+       * predkeys[i]; activeOutcomes[id] = oid; activeParams[id] =
+       * PARAMS[pid].getParams(oid); id++; }
+       */
+      sortPreds[pid] = new ComparablePredicate(PRED_LABELS[pid],
+          predkeys, activeParams);
     }
-    System.err.println("Compressed " + PARAMS.length + " parameters to " + numPreds);
-    sortPreds = new ComparablePredicate[numPreds];
-    System.arraycopy(tmpPreds, 0, sortPreds, 0, numPreds);
+
     Arrays.sort(sortPreds);
     return sortPreds;
   }
 
+  protected List<List<ComparablePredicate>> compressOutcomes(ComparablePredicate[] sorted) {
+    List<List<ComparablePredicate>> outcomePatterns = new ArrayList<>();
+    if (sorted.length > 0) {
+      ComparablePredicate cp = sorted[0];
+      List<ComparablePredicate> newGroup = new ArrayList<>();
+      for (int i = 0; i < sorted.length; i++) {
+        if (cp.compareTo(sorted[i]) == 0) {
+          newGroup.add(sorted[i]);
+        } else {
+          cp = sorted[i];
+          outcomePatterns.add(newGroup);
+          newGroup = new ArrayList<>();
+          newGroup.add(sorted[i]);
+        }
+      }
+      outcomePatterns.add(newGroup);
+    }
+    return outcomePatterns;
+  }
+
+
 
   protected List<List<ComparablePredicate>> computeOutcomePatterns(ComparablePredicate[] sorted) {
     ComparablePredicate cp = sorted[0];

http://git-wip-us.apache.org/repos/asf/opennlp/blob/15862426/opennlp-tools/src/test/java/opennlp/tools/ml/naivebayes/NaiveBayesSerializedCorrectnessTest.java
----------------------------------------------------------------------
diff --git a/opennlp-tools/src/test/java/opennlp/tools/ml/naivebayes/NaiveBayesSerializedCorrectnessTest.java b/opennlp-tools/src/test/java/opennlp/tools/ml/naivebayes/NaiveBayesSerializedCorrectnessTest.java
new file mode 100644
index 0000000..0146885
--- /dev/null
+++ b/opennlp-tools/src/test/java/opennlp/tools/ml/naivebayes/NaiveBayesSerializedCorrectnessTest.java
@@ -0,0 +1,184 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package opennlp.tools.ml.naivebayes;
+
+import java.io.BufferedReader;
+import java.io.BufferedWriter;
+import java.io.File;
+import java.io.IOException;
+import java.io.StringReader;
+import java.io.StringWriter;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.HashMap;
+
+import org.junit.Assert;
+import org.junit.Before;
+import org.junit.Test;
+
+import opennlp.tools.ml.AbstractTrainer;
+import opennlp.tools.ml.model.AbstractDataIndexer;
+import opennlp.tools.ml.model.DataIndexer;
+import opennlp.tools.ml.model.Event;
+import opennlp.tools.ml.model.TwoPassDataIndexer;
+import opennlp.tools.util.TrainingParameters;
+
+/**
+ * Test for naive bayes classification correctness without smoothing
+ */
+public class NaiveBayesSerializedCorrectnessTest {
+
+  private DataIndexer testDataIndexer;
+
+  @Before
+  public void initIndexer() {
+    TrainingParameters trainingParameters = new TrainingParameters();
+    trainingParameters.put(AbstractTrainer.CUTOFF_PARAM, "1");
+    trainingParameters.put(AbstractDataIndexer.SORT_PARAM, "false");;
+    testDataIndexer = new TwoPassDataIndexer();
+    testDataIndexer.init(trainingParameters, new HashMap<>());
+  }
+
+  @Test
+  public void testNaiveBayes1() throws IOException {
+
+    testDataIndexer.index(NaiveBayesCorrectnessTest.createTrainingStream());
+    NaiveBayesModel model1 =
+        (NaiveBayesModel) new NaiveBayesTrainer().trainModel(testDataIndexer);
+
+    NaiveBayesModel model2 = persistedModel(model1);
+
+    String label = "politics";
+    String[] context = {"bow=united", "bow=nations"};
+    Event event = new Event(label, context);
+
+    testModelOutcome(model1, model2, event);
+
+  }
+
+  @Test
+  public void testNaiveBayes2() throws IOException {
+
+    testDataIndexer.index(NaiveBayesCorrectnessTest.createTrainingStream());
+    NaiveBayesModel model1 =
+        (NaiveBayesModel) new NaiveBayesTrainer().trainModel(testDataIndexer);
+
+    NaiveBayesModel model2 = persistedModel(model1);
+
+    String label = "sports";
+    String[] context = {"bow=manchester", "bow=united"};
+    Event event = new Event(label, context);
+
+    testModelOutcome(model1, model2, event);
+
+  }
+
+  @Test
+  public void testNaiveBayes3() throws IOException {
+
+    testDataIndexer.index(NaiveBayesCorrectnessTest.createTrainingStream());
+    NaiveBayesModel model1 =
+        (NaiveBayesModel) new NaiveBayesTrainer().trainModel(testDataIndexer);
+
+    NaiveBayesModel model2 = persistedModel(model1);
+
+    String label = "politics";
+    String[] context = {"bow=united"};
+    Event event = new Event(label, context);
+
+    testModelOutcome(model1, model2, event);
+
+  }
+
+  @Test
+  public void testNaiveBayes4() throws IOException {
+
+    testDataIndexer.index(NaiveBayesCorrectnessTest.createTrainingStream());
+    NaiveBayesModel model1 =
+        (NaiveBayesModel) new NaiveBayesTrainer().trainModel(testDataIndexer);
+
+    NaiveBayesModel model2 = persistedModel(model1);
+
+    String label = "politics";
+    String[] context = {};
+    Event event = new Event(label, context);
+
+    testModelOutcome(model1, model2, event);
+
+  }
+
+
+  @Test
+  public void testPlainTextModel() throws IOException {
+    testDataIndexer.index(NaiveBayesCorrectnessTest.createTrainingStream());
+    NaiveBayesModel model1 =
+        (NaiveBayesModel) new NaiveBayesTrainer().trainModel(testDataIndexer);
+
+
+    StringWriter sw1 = new StringWriter();
+
+    NaiveBayesModelWriter modelWriter =
+        new PlainTextNaiveBayesModelWriter(model1, new BufferedWriter(sw1));
+    modelWriter.persist();
+
+    NaiveBayesModelReader reader =
+        new PlainTextNaiveBayesModelReader(new BufferedReader(new StringReader(sw1.toString())));
+    reader.checkModelType();
+
+    NaiveBayesModel model2 = (NaiveBayesModel)reader.constructModel();
+
+    StringWriter sw2 = new StringWriter();
+    modelWriter = new PlainTextNaiveBayesModelWriter(model2, new BufferedWriter(sw2));
+    modelWriter.persist();
+
+    System.out.println(sw1.toString());
+    Assert.assertEquals(sw1.toString(), sw2.toString());
+
+  }
+
+  protected static NaiveBayesModel persistedModel(NaiveBayesModel model) throws IOException {
+    Path tempFilePath = Files.createTempFile("ptnb-", ".bin");
+    File file = tempFilePath.toFile();
+    NaiveBayesModelWriter modelWriter = new BinaryNaiveBayesModelWriter(model, tempFilePath.toFile());
+    modelWriter.persist();
+    NaiveBayesModelReader reader = new BinaryNaiveBayesModelReader(file);
+    reader.checkModelType();
+    return (NaiveBayesModel)reader.constructModel();
+  }
+
+  protected static void testModelOutcome(NaiveBayesModel model1, NaiveBayesModel model2, Event event) {
+    String[] labels1 = extractLabels(model1);
+    String[] labels2 = extractLabels(model2);
+
+    Assert.assertArrayEquals(labels1, labels2);
+
+    double[] outcomes1 = model1.eval(event.getContext());
+    double[] outcomes2 = model2.eval(event.getContext());
+
+    Assert.assertArrayEquals(outcomes1, outcomes2, 0.000000000001);
+
+  }
+
+  private static String[] extractLabels(NaiveBayesModel model) {
+    String[] labels = new String[model.getNumOutcomes()];
+    for (int i = 0; i < model.getNumOutcomes(); i++) {
+      labels[i] = model.getOutcome(i);
+    }
+    return labels;
+  }
+}