Emanuel Huber commited on
Commit
b29b5d8
1 Parent(s): 7874b31

Added confidence scores

Browse files
Files changed (1) hide show
  1. app.py +8 -2
app.py CHANGED
@@ -5,6 +5,7 @@ from typing import List, Tuple
5
  import gradio as gr
6
  import pandas as pd
7
  import spacy
 
8
  from transformers import AutoModelForTokenClassification, AutoTokenizer
9
 
10
  try:
@@ -36,6 +37,7 @@ def predict(text, nlp, logger=None) -> Tuple[List[str], List[str]]:
36
 
37
  i_token = 0
38
  labels = []
 
39
  for off, is_special_token, pred in zip(
40
  input_tokens["offset_mapping"][0],
41
  input_tokens["special_tokens_mask"][0],
@@ -47,17 +49,21 @@ def predict(text, nlp, logger=None) -> Tuple[List[str], List[str]]:
47
  if logger is not None:
48
  logger.info("{}, {}, {}".format(off, tokens[i_token], label))
49
  labels.append(label)
 
 
 
50
  i_token += 1
51
 
52
- return tokens, labels
53
 
54
 
55
  def text_analysis(text):
56
- tokens, labels = predict(text, nlp, logger)
57
  pos_count = pd.DataFrame(
58
  {
59
  "token": tokens,
60
  "etiqueta": labels,
 
61
  }
62
  )
63
  pos_tokens = []
 
5
  import gradio as gr
6
  import pandas as pd
7
  import spacy
8
+ import torch
9
  from transformers import AutoModelForTokenClassification, AutoTokenizer
10
 
11
  try:
 
37
 
38
  i_token = 0
39
  labels = []
40
+ scores = []
41
  for off, is_special_token, pred in zip(
42
  input_tokens["offset_mapping"][0],
43
  input_tokens["special_tokens_mask"][0],
 
49
  if logger is not None:
50
  logger.info("{}, {}, {}".format(off, tokens[i_token], label))
51
  labels.append(label)
52
+ scores.append(
53
+ "{:.2f}".format(100 * float(torch.softmax(pred, dim=-1).detach().max()))
54
+ )
55
  i_token += 1
56
 
57
+ return tokens, labels, scores
58
 
59
 
60
  def text_analysis(text):
61
+ tokens, labels, scores = predict(text, nlp, logger)
62
  pos_count = pd.DataFrame(
63
  {
64
  "token": tokens,
65
  "etiqueta": labels,
66
+ "confiança": scores,
67
  }
68
  )
69
  pos_tokens = []