DD0101 commited on
Commit
8e41965
1 Parent(s): 8a2ba57

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +8 -4
app.py CHANGED
@@ -73,7 +73,6 @@ parser.add_argument("--model_dir", default="/home/user/app/Customized_IDSF/Joint
73
  parser.add_argument("--batch_size", default=32, type=int, help="Batch size for prediction")
74
  parser.add_argument("--no_cuda", action="store_true", help="Avoid using CUDA when available")
75
 
76
- globals()['var'] = "an object"
77
  pred_config = parser.parse_args()
78
 
79
  # load model and args
@@ -124,17 +123,22 @@ def ner(text):
124
 
125
  slot_tokens = []
126
 
127
- words[0].capitalize()
128
  for word, pred in zip(words, slot_preds):
129
  word = word.replace("_", " ")
130
  if pred == 'O':
131
  slot_tokens.extend([(word, None), (" ", None)])
132
- elif pred[0] == 'I':
 
133
  added_tokens = list(slot_tokens[-2])
134
  added_tokens[0] += f' {word}'
135
  slot_tokens[-2] = tuple(added_tokens)
136
- else:
 
137
  slot_tokens.extend([(word, pred[2:]), (" ", None)])
 
 
 
138
 
139
  intent_label = intent_label_lst[intent_pred]
140
 
 
73
  parser.add_argument("--batch_size", default=32, type=int, help="Batch size for prediction")
74
  parser.add_argument("--no_cuda", action="store_true", help="Avoid using CUDA when available")
75
 
 
76
  pred_config = parser.parse_args()
77
 
78
  # load model and args
 
123
 
124
  slot_tokens = []
125
 
126
+ words[0] = words[0].upper() + words[1:]
127
  for word, pred in zip(words, slot_preds):
128
  word = word.replace("_", " ")
129
  if pred == 'O':
130
  slot_tokens.extend([(word, None), (" ", None)])
131
+
132
+ elif pred[0] == 'I': # token with label's prefix' "I-XXX": will be combined their corresponded "B-XXX"
133
  added_tokens = list(slot_tokens[-2])
134
  added_tokens[0] += f' {word}'
135
  slot_tokens[-2] = tuple(added_tokens)
136
+
137
+ elif pred[0] == 'B': # token with label's prefix "B-XXX" : remove "B-"
138
  slot_tokens.extend([(word, pred[2:]), (" ", None)])
139
+
140
+ else: # PAD or UNK tags
141
+ slot_tokens.extend([(word, pred), (" ", None)])
142
 
143
  intent_label = intent_label_lst[intent_pred]
144