Add doc and adjust some codes.

data_utils/data.py

@@ -103,8 +103,8 @@ class DataGenerator(object):
         :type filename: basestring | file
         :param transcript: Transcription text.
         :type transcript: basestring
-        :return: Tuple of audio feature tensor and list of token ids for
+        :return: Tuple of audio feature tensor and data of transcription part,
-                 transcription.
+                 where transcription part could be token ids or text.
         :rtype: tuple of (2darray, list)
         """
         if filename.startswith('tar:'):

infer.py

@@ -103,9 +103,7 @@ def infer():
         num_processes=args.num_proc_bsearch)
 
     error_rate_func = cer if args.error_rate_type == 'cer' else wer
-    target_transcripts = [
+    target_transcripts = [transcript for _, transcript in infer_data]
-        transcript for _, transcript in infer_data
-    ]
     for target, result in zip(target_transcripts, result_transcripts):
         print("\nTarget Transcription: %s\nOutput Transcription: %s" %
               (target, result))

test.py

@@ -104,9 +104,7 @@ def evaluate():
             vocab_list=vocab_list,
             language_model_path=args.lang_model_path,
             num_processes=args.num_proc_bsearch)
-        target_transcripts = [
+        target_transcripts = [transcript for _, transcript in infer_data]
-            transcript for _, transcript in infer_data
-        ]
         for target, result in zip(target_transcripts, result_transcripts):
             error_sum += error_rate_func(target, result)
             num_ins += 1

tools/tune.py

@@ -164,9 +164,7 @@ def tune():
             for i in xrange(len(infer_data))
         ]
 
-        target_transcripts = [
+        target_transcripts = [transcript for _, transcript in infer_data]
-            transcript for _, transcript in infer_data
-        ]
 
         num_ins += len(target_transcripts)
         # grid search