sandbox/language_id: code and script updates for evaluating the model.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/language_id@3639 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

egs/lre/v1/run_logistic_regression.sh

@@ -29,9 +29,25 @@ posterior_output=posteriors
 scores=posteriors
 
 classes=exp/ivectors_train/trials
-utils/utt2lang_to_utt2langint.py exp/ivectors_train/languages.txt \
-    data/train/utt2lang $trials
 
-logistic-regression-eval $model scp:$train_ivectors $posterior_output 2>$log
-logistic-regression-eval $model ark:$trials scp:$train_ivectors $scores 2>$log
+trials="utils/sym2int.pl -f 2 exp/ivectors_train/languages.txt data/train/utt2lang|"
+scores="|utils/int2sym.pl -f 2 exp/ivectors_train/languages.txt  >exp/ivectors_train/train_scores"
 
+logistic-regression-eval $model scp:$train_ivectors ark,t:exp/ivectors_train/posteriors
+logistic-regression-eval $model "ark:$trials" scp:$train_ivectors "$scores"
+
+logistic-regression-eval $model scp:$train_ivectors ark,t:- | \
+  awk '{max=$3; argmax=3; for(f=3;f<NF;f++) { if ($f>max) { max=$f; argmax=f; }}  print $1, (argmax - 3); }' | \
+  utils/int2sym.pl -f 2 exp/ivectors_train/languages.txt >exp/ivectors_train/train_output
+
+# note: we treat the language as a sentence; it happens that the WER/SER
+# corresponds to the recognition error rate.
+compute-wer --text ark:data/train/utt2lang ark:exp/ivectors_train/train_output
+
+# It perfectly classifies the training data:
+#%WER 0.00 [ 0 / 10173, 0 ins, 0 del, 0 sub ]
+#%SER 0.00 [ 0 / 10173 ]
+#Scored 10173 sentences, 0 not present in hyp.
+
+
+# someone needs to extend this to run on the dev data.

src/ivectorbin/logistic-regression-eval.cc

@@ -24,16 +24,17 @@
 
 using namespace kaldi;
 
-void posteriors(ParseOptions &po, const LogisticRegressionConfig &config) {
+int ComputePosteriors(ParseOptions &po, const LogisticRegressionConfig &config) {
   std::string model = po.GetArg(1),
-        vector_rspecifier = po.GetArg(2),
-        posteriors_out = po.GetArg(3);
-
-  LogisticRegression classifier = LogisticRegression();
+      vector_rspecifier = po.GetArg(2),
+      posteriors_wspecifier = po.GetArg(3);
+  
+  LogisticRegression classifier;
   ReadKaldiObject(model, &classifier);
   
   std::vector<Vector<BaseFloat> > vectors;
   SequentialBaseFloatVectorReader vector_reader(vector_rspecifier);
+  BaseFloatVectorWriter posterior_writer(posteriors_wspecifier);
   std::vector<std::string> utt_list;
   int32 num_utt_done = 0;
 
@@ -45,6 +46,11 @@ void posteriors(ParseOptions &po, const LogisticRegressionConfig &config) {
     num_utt_done++;
   }
 
+  if (vectors.empty()) {
+    KALDI_WARN << "Read no input";
+    return 1;
+  }
+  
   Matrix<double> xs(vectors.size(), vectors[0].Dim());
   for (int i = 0; i < vectors.size(); i++) {
     xs.Row(i).CopyFromVec(vectors[i]);
@@ -52,19 +58,20 @@ void posteriors(ParseOptions &po, const LogisticRegressionConfig &config) {
  
   Matrix<double> posteriors;
   classifier.GetPosteriors(xs, &posteriors);
-  std::ofstream posteriors_xwstream(posteriors_out.c_str(), std::ios::out);
+  
   KALDI_LOG << "Calculated posteriors for " << num_utt_done << " vectors.";
   for (int i = 0; i < posteriors.NumRows(); i++) {
-    posteriors_xwstream << utt_list[i] << " " << posteriors.Row(i);
-  }
-  posteriors_xwstream.close();
+    Vector<BaseFloat> row(posteriors.Row(i));
+    posterior_writer.Write(utt_list[i], row);
+  }  
+  return (num_utt_done == 0 ? 1 : 0);
 }
 
-void scores(ParseOptions &po, const LogisticRegressionConfig &config) {
+int32 ComputeScores(ParseOptions &po, const LogisticRegressionConfig &config) {
   std::string model_rspecifier = po.GetArg(1),
-        trials_rspecifier = po.GetArg(2),
-        vector_rspecifier = po.GetArg(3),
-        scores_out = po.GetArg(4);
+      trials_rspecifier = po.GetArg(2),
+      vector_rspecifier = po.GetArg(3),
+      scores_out = po.GetArg(4);
 
   SequentialInt32Reader class_reader(trials_rspecifier);
   LogisticRegression classifier = LogisticRegression();
@@ -91,6 +98,11 @@ void scores(ParseOptions &po, const LogisticRegressionConfig &config) {
     }
   }
 
+  if (vectors.empty()) {
+    KALDI_WARN << "Read no input";
+    return 1;
+  }
+  
   Matrix<double> xs(vectors.size(), vectors[0].Dim());
   for (int i = 0; i < vectors.size(); i++) {
     xs.Row(i).CopyFromVec(vectors[i]);
@@ -98,15 +110,17 @@ void scores(ParseOptions &po, const LogisticRegressionConfig &config) {
  
   Matrix<double> posteriors;
   classifier.GetPosteriors(xs, &posteriors);
+
+  bool binary = false;
+  Output ko(scores_out.c_str(), binary);
   
-  std::ofstream scores_xwstream(scores_out.c_str(), std::ios::out);
   for (int i = 0; i < ys.size(); i++) {
-    scores_xwstream << utt_list[i] << " " << ys[i] << " " << posteriors(i, ys[i]) << std::endl;
+    ko.Stream() << utt_list[i] << " " << ys[i] << " " << posteriors(i, ys[i]) << std::endl;
   }
-  scores_xwstream.close();
-  KALDI_LOG << "Calculated scores for" << num_utt_done 
+  KALDI_LOG << "Calculated scores for " << num_utt_done 
             << " vectors with "
             << num_utt_err << " missing. ";
+  return (num_utt_done == 0 ? 1 : 0);
 }
 
 int main(int argc, char *argv[]) {
@@ -116,10 +130,10 @@ int main(int argc, char *argv[]) {
     const char *usage =
         "Evaluates a model on input vectors and outputs either\n"
         "posterior probability or scores.\n"
-        "Usage1: logistic-regression-eval <model> <input-vectors>\n"
-        "                                <output-posteriors>\n"
-        "Usage2: logistic-regression-eval <model> <trials> <input-vectors>\n"
-        "                                <output-scores>\n";
+        "Usage1: logistic-regression-eval <model> <input-vectors-rspecifier>\n"
+        "                                <output-posteriors-wspecifier>\n"
+        "Usage2: logistic-regression-eval <model> <trials-file> <input-vectors-rspecifier>\n"
+        "                                <output-scores-file>\n";
     
   ParseOptions po(usage);
 
@@ -134,9 +148,9 @@ int main(int argc, char *argv[]) {
     exit(1);
   }
   
-  (po.NumArgs() == 4) ? scores(po, config) : posteriors(po, config);
-
-  return 0;
+  return (po.NumArgs() == 4) ?
+      ComputeScores(po, config) :
+      ComputePosteriors(po, config);
   } catch(const std::exception &e) {
     std::cerr << e.what();
     return -1;