lisaterumi commited on
Commit
a0158e4
1 Parent(s): 59f6643

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +99 -0
README.md ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: "pt"
3
+ widget:
4
+ - text: "Dispneia importante aos esforços + dor tipo peso no peito no esforço."
5
+ - text: "Obeso, has, icc c # cintilografia miocardica para avaliar angina. Discreto edema mmii pricn a esquerda."
6
+ - text: "Plastia Mitral ( Insuficiencia ), CRM Saf-2Mg e e Saf-3MG ).(09/03/16). Nega palpitação."
7
+ - text: "Uso: AAS 100 -1xd; Metoprolol 25 -1xd; FSM -1xd ; Levotiroxina 175 -1xd; Sinva 40 -1xd; Fluoxetina 20-1xd."
8
+ - text: "Refere melhora da dispneia depois da cx porem mantem aos mdoeardos-leves esforço."
9
+
10
+ datasets:
11
+ - TempClinBr
12
+ ---
13
+
14
+ # Portuguese NER- TempClinBr - BioBERTpt(all)
15
+
16
+ Treinado com BioBERTpt(all), com o corpus TempClinBr.
17
+
18
+ Metricas:
19
+
20
+ ```
21
+ precision recall f1-score support
22
+
23
+ 0 0.75 0.90 0.82 291
24
+ 1 0.77 1.00 0.87 33
25
+ 2 1.00 0.25 0.40 28
26
+ 3 0.90 0.99 0.94 71
27
+ 4 0.79 0.91 0.85 112
28
+ 5 0.72 0.83 0.77 420
29
+ 6 0.62 0.45 0.53 11
30
+ 7 0.96 0.85 0.91 2236
31
+ 8 0.61 0.67 0.64 78
32
+ 9 0.61 0.98 0.76 124
33
+ 10 0.81 0.87 0.84 503
34
+ 11 0.67 0.60 0.63 10
35
+
36
+ accuracy 0.86 3917
37
+ macro avg 0.77 0.78 0.74 3917
38
+ weighted avg 0.87 0.86 0.86 3917
39
+
40
+ F1: 0.8588744393393593 Accuracy: 0.8565228491192239
41
+
42
+ ```
43
+
44
+ Parâmetros:
45
+
46
+ ```
47
+ device = cuda (Colab)
48
+ nclasses = len(tag2id)
49
+ nepochs = 50 => parou na 9
50
+ batch_size = 16
51
+ batch_status = 32
52
+ learning_rate = 3e-5
53
+
54
+ early_stop = 5
55
+ max_length = 256
56
+ write_path = 'model'
57
+ ```
58
+
59
+ Eval no conjunto de teste - TempClinBr
60
+ OBS: Avaliação com tag "O" (label 7), se necessário fazer a média sem essa tag.
61
+
62
+ ```
63
+ tag2id ={'B-Tratamento': 0,
64
+ 'I-Teste': 1,
65
+ 'I-Ocorrencia': 2,
66
+ 'B-Evidencia': 3,
67
+ 'B-Teste': 4,
68
+ 'I-Problema': 5,
69
+ 'B-DepartamentoClinico': 6,
70
+ 'O': 7,
71
+ 'I-Tratamento': 8,
72
+ 'B-Ocorrencia': 9,
73
+ 'B-Problema': 10,
74
+ 'I-DepartamentoClinico': 11,
75
+ '<pad>': 12}
76
+
77
+ precision recall f1-score support
78
+
79
+ 0 0.82 0.92 0.87 261
80
+ 1 0.81 0.58 0.67 99
81
+ 2 0.56 0.20 0.29 51
82
+ 3 1.00 0.94 0.97 128
83
+ 4 0.81 0.86 0.83 194
84
+ 5 0.81 0.87 0.84 645
85
+ 6 0.96 0.80 0.87 30
86
+ 7 0.95 0.90 0.93 2431
87
+ 8 0.73 0.81 0.77 146
88
+ 9 0.74 0.88 0.80 146
89
+ 10 0.87 0.95 0.91 713
90
+ 11 0.83 0.71 0.77 14
91
+ 12 0.00 0.00 0.00 0
92
+
93
+ accuracy 0.89 4858
94
+ macro avg 0.76 0.72 0.73 4858
95
+ weighted avg 0.89 0.89 0.89 4858
96
+ ```
97
+
98
+
99
+ Como citar: **em breve**