Package cc.mallet.types

Examples of cc.mallet.types.LabelAlphabet

59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
  public void evaluate (String description, Extraction extraction, PrintStream out)
  {
    int numDocs = extraction.getNumDocuments ();
    assert numDocs == extraction.getNumRecords ();

    LabelAlphabet dict = extraction.getLabelAlphabet();
    int numLabels = dict.size();
    int[] numCorr = new int [numLabels];
    int[] numPred = new int [numLabels];
    int[] numTrue = new int [numLabels];

    for (int docnum = 0; docnum < numDocs; docnum++) {
      Record extracted = extraction.getRecord (docnum);
      Record target = extraction.getTargetRecord (docnum);

      // Calc precision
      Iterator it = extracted.fieldsIterator ();
      while (it.hasNext ()) {
        Field predField = (Field) it.next ();
        Label name = predField.getName ();
        Field trueField = target.getField (name);
        int idx = name.getIndex ();

        for (int j = 0; j < predField.numValues(); j++) {
          numPred [idx]++;
          if (trueField != null && trueField.isValue (predField.value (j), comparator)) {
            numCorr [idx]++;
          } else {
            // We have an error, report if necessary (this should be moved to the per-field rather than per-filler level.)
            if (errorOutputStream != null) {
              //xxx TODO: Display name of supporting document
              errorOutputStream.println ("Error in extraction!");
              errorOutputStream.println ("Predicted "+predField);
              errorOutputStream.println ("True "+trueField);
              errorOutputStream.println ();
            }
          }

        }
      }

      // Calc true
      it = target.fieldsIterator ();
      while (it.hasNext ()) {
        Field trueField = (Field) it.next ();
        Label name = trueField.getName ();
        numTrue [name.getIndex ()] += trueField.numValues ();
      }
    }

    out.println (description+" SEGMENT counts");
    out.println ("Name\tCorrect\tPred\tTarget");
    for (int i = 0; i < numLabels; i++) {
      Label name = dict.lookupLabel (i);
      out.println (name+"\t"+numCorr[i]+"\t"+numPred[i]+"\t"+numTrue[i]);
    }
    out.println ();

    DecimalFormat f = new DecimalFormat ("0.####");

    double totalF1 = 0;
    int totalFields = 0;
    out.println (description+" per-field F1");
    out.println ("Name\tP\tR\tF1");
    for (int i = 0; i < numLabels; i++) {
      double P = (numPred[i] == 0) ? 0 : ((double)numCorr[i]) / numPred [i];
      double R = (numTrue[i] == 0) ? 1 : ((double)numCorr[i]) / numTrue [i];
      double F1 = (P + R == 0) ? 0 : (2 * P * R) / (P + R);
      if ((numPred[i] > 0) || (numTrue[i] > 0)) {
        totalF1 += F1;
        totalFields++;
      }
      Label name = dict.lookupLabel (i);
      out.println (name+"\t"+f.format(P)+"\t"+f.format(R)+"\t"+f.format(F1));
    }

    int totalCorr = MatrixOps.sum (numCorr);
    int totalPred = MatrixOps.sum (numPred);
View Full Code Here
73
74
75
76
77
78
79
80
81
82
83
84
85

  private static LabelAlphabet createBlankAlphabet (int numOutcomes)
  {
    if (numOutcomes > 0) {
      LabelAlphabet outcomes = new LabelAlphabet ();
      /* Setup default outcomes */
      for (int i = 0; i < numOutcomes; i++) {
        outcomes.lookupIndex (new Integer (i));
      }
      return outcomes;
    } else return null;
  }
View Full Code Here
29
30
31
32
33
34
35
36
37
38
39
  private LabelAlphabet ld;
  private LabelVector lv;

  protected void setUp ()
  {
    ld = new LabelAlphabet ();
    lv = new LabelVector (ld,
                          new int[] {
                            ld.lookupIndex ("a"),
                            ld.lookupIndex ("b"),
                            ld.lookupIndex ("c"),
View Full Code Here
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
  public void evaluate (String description, Extraction extraction, PrintWriter out)
  {
    int numDocs = extraction.getNumDocuments ();
    assert numDocs == extraction.getNumRecords ();

    LabelAlphabet dict = extraction.getLabelAlphabet();
    int numLabels = dict.size();
    int[] numCorr = new int [numLabels];
    int[] numPred = new int [numLabels];
    int[] numTrue = new int [numLabels];

    for (int docnum = 0; docnum < numDocs; docnum++) {
      Record extracted = extraction.getRecord (docnum);
      Record target = extraction.getTargetRecord (docnum);

      // Calc precision
      Iterator it = extracted.fieldsIterator ();
      while (it.hasNext ()) {
        Field predField = (Field) it.next ();
        Label name = predField.getName ();
        Field trueField = target.getField (name);
        int idx = name.getIndex ();

        numPred [idx]++;

        if (predField.numValues() > 1)
          System.err.println ("Warning: Field "+predField+" has more than one extracted value. Picking arbitrarily...");
        if (trueField != null && trueField.isValue (predField.value (0), comparator)) {
          numCorr [idx]++;
        } else {
          // We have an error, report if necessary
          if (errorOutputStream != null) {
            //xxx TODO: Display name of supporting document
            errorOutputStream.println ("Error in extraction! Document "+extraction.getDocumentExtraction (docnum).getName ());
            errorOutputStream.println ("Predicted "+predField);
            errorOutputStream.println ("True "+trueField);
            errorOutputStream.println ();
          }
        }
      }

      // Calc true
      it = target.fieldsIterator ();
      while (it.hasNext ()) {
        Field trueField = (Field) it.next ();
        Label name = trueField.getName ();
        numTrue [name.getIndex ()]++;
      }
    }

    DecimalFormat f = new DecimalFormat ("0.####");

    double totalF1 = 0;
    int totalFields = 0;
    out.println (description+" per-document F1");
    out.println ("Name\tP\tR\tF1");
    for (int i = 0; i < numLabels; i++) {
      double P = (numPred[i] == 0) ? 0 : ((double)numCorr[i]) / numPred [i];
      double R = (numTrue[i] == 0) ? 1 : ((double)numCorr[i]) / numTrue [i];
      double F1 = (P + R == 0) ? 0 : (2 * P * R) / (P + R);
      if ((numPred[i] > 0) || (numTrue[i] > 0)) {
        totalF1 += F1;
        totalFields++;
      }
      Label name = dict.lookupLabel (i);
      out.println (name+"\t"+f.format(P)+"\t"+f.format(R)+"\t"+f.format(F1));
    }

    int totalCorr = MatrixOps.sum (numCorr);
    int totalPred = MatrixOps.sum (numPred);
View Full Code Here
121
122
123
124
125
126
127
128
129
130
131
132
          maxidx = assn.get (var);
        } else {
          maxidx = 0;
        }

        LabelAlphabet dict = labelOfVar (var).getLabelAlphabet ();
        theseLabels[i] = dict.lookupLabel (maxidx);
      }

      lbls[t] = new Labels (theseLabels);
    }
View Full Code Here
74
75
76
77
78
79
80
81
82
83
84
    out.println (".pred { background-color:#FFFF66 }");
    out.close ();


    //PRED css
    LabelAlphabet dict = extraction.getLabelAlphabet ();
    String[] fields = determineFieldNames (dict);
    String[] colors = ColorUtils.rainbow (fields.length, (float) SATURATION, 1);
    out = new PrintWriter (new FileWriter (new File (directory, DOC_ERRS_PRED_CSS_FNAME)));
    out.println (".class_legend { border-style: dashed; border-width: 2px; padding: 10px; padding-top: 0ex; float: right; margin:2em; }");
    out.println (".tf_legend { visibility: hidden; }");
View Full Code Here
142
143
144
145
146
147
148
  private class OverlappingFeaturePipe extends Pipe {

    private static final long serialVersionUID = 1L;

    public OverlappingFeaturePipe () {
      super (new Alphabet(), new LabelAlphabet());     
    }
View Full Code Here
186
187
188
189
190
191
192
193
194
195
            positive = false;
            break;
          }
        }
      }
      LabelAlphabet ldict = (LabelAlphabet)getTargetAlphabet();
      String label = positive ? "YES" : "NO";     
      carrier.setTarget(ldict.lookupLabel(label));
      return carrier;
    }
View Full Code Here
476
477
478
479
480
481
482
483
484
485
486
487
488
        char type = in.readChar ();
        Object obj;
       
        switch (type) {
        case TYPE_LABEL:
            LabelAlphabet ldict = (LabelAlphabet) getTargetAlphabet ();
            String name = (String) in.readObject ();
            obj = ldict.lookupLabel (name);
            break;
        case TYPE_FEATURE_VECTOR:
            int[] indices = (int[]) in.readObject ();
            double[] values = (double[]) in.readObject ();
            obj = new FeatureVector(getDataAlphabet (), indices, values);
View Full Code Here
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
    super (name);
  }

  public void testSerializable () throws IOException, ClassNotFoundException
  {
    LabelAlphabet dict = new LabelAlphabet ();
    Labels lbls1 = new Labels (new Label[] {
      dict.lookupLabel ("A"),
      dict.lookupLabel ("B"),
    });
    Labels lbls2 = new Labels (new Label[] {
      dict.lookupLabel ("C"),
      dict.lookupLabel ("A"),
    });
    LabelsSequence lblseq  = new LabelsSequence (new Labels[] { lbls1, lbls2 });
    LabelsSequence lblseq2 = (LabelsSequence) TestSerializable.cloneViaSerialization (lblseq);
    assertEquals (lblseq.size(), lblseq2.size());
    assertEquals (lblseq.getLabels(0).toString(), lblseq2.getLabels(0).toString ());
View Full Code Here
TOP

Related Classes of cc.mallet.types.LabelAlphabet

  • cc.mallet.classify.evaluate.ConfusionMatrix
  • cc.mallet.classify.MaxEnt
  • cc.mallet.classify.MaxEntOptimizableByLabelDistribution
  • cc.mallet.classify.MaxEntOptimizableByLabelLikelihood
  • cc.mallet.classify.MCMaxEnt
  • cc.mallet.classify.MCMaxEntTrainer$MaximizableTrainer
  • cc.mallet.classify.RankMaxEnt
  • cc.mallet.classify.RankMaxEntTrainer$MaximizableTrainer
  • cc.mallet.cluster.examples.FirstOrderClusterExample$OverlappingFeaturePipe
  • cc.mallet.cluster.tui.Clusterings2Clusterer$ClusteringPipe
    • Copyright © 2018 www.massapicom. 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.