-
Notifications
You must be signed in to change notification settings - Fork 0
/
named_entity_recognition.txt
123 lines (123 loc) · 8.21 KB
/
named_entity_recognition.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
-- AIT726 Homework 3 from Julia Jeng, Shu Wang, and Arman Anwar --
[Info] Load 14041 training sentences (max:113 words) from .//conll2003//train.txt.
[Info] Load 3250 validation sentences (max:109 words) from .//conll2003//valid.txt.
[Info] Load 3453 testing sentences (max:124 words) from .//conll2003//test.txt.
[Info] Get 28449 vocabulary words successfully.
[Info] Load pre-trained word2vec weights from .//temp//preWeights.npy.
[Demo] --- RNNType: RNN | HiddenNodes: 256 | Bi-Direction: False | Pre-Trained: True ---
[Para] BatchSize=256, LearningRate=0.0010, MaxEpoch=1000, PerEpoch=5.
[Epoch 005] loss: 0.826, train acc: 85.768%, valid acc: 85.302%.
[Epoch 010] loss: 0.620, train acc: 88.201%, valid acc: 86.311%.
[Epoch 015] loss: 0.466, train acc: 90.698%, valid acc: 87.757%.
[Epoch 020] loss: 0.360, train acc: 92.490%, valid acc: 89.644%.
[Eval] Testing accuracy: 87.361%.
processed 46435 tokens with 5648 phrases; found: 7268 phrases; correct: 2983.
accuracy: 58.11%; (non-O)
accuracy: 87.36%; precision: 41.04%; recall: 52.82%; FB1: 46.19
LOC: precision: 53.19%; recall: 69.48%; FB1: 60.25 2179
MISC: precision: 29.83%; recall: 40.03%; FB1: 34.18 942
ORG: precision: 31.15%; recall: 48.77%; FB1: 38.02 2600
PER: precision: 47.38%; recall: 45.33%; FB1: 46.33 1547
[Info] Save the RNN model in .//models//model_RNN_preTrue_biFalse_256_256_0.001.pth
[Info] --------------------------------------------------------------------------------
[Demo] --- RNNType: RNN | HiddenNodes: 256 | Bi-Direction: True | Pre-Trained: True ---
[Para] BatchSize=256, LearningRate=0.0010, MaxEpoch=1000, PerEpoch=5.
[Epoch 005] loss: 0.660, train acc: 87.834%, valid acc: 86.029%.
[Epoch 010] loss: 0.438, train acc: 90.745%, valid acc: 87.699%.
[Epoch 015] loss: 0.287, train acc: 93.302%, valid acc: 88.869%.
[Eval] Testing accuracy: 86.876%.
processed 46435 tokens with 5648 phrases; found: 7504 phrases; correct: 3105.
accuracy: 63.28%; (non-O)
accuracy: 86.88%; precision: 41.38%; recall: 54.98%; FB1: 47.22
LOC: precision: 54.73%; recall: 68.65%; FB1: 60.90 2092
MISC: precision: 29.61%; recall: 40.03%; FB1: 34.04 949
ORG: precision: 30.92%; recall: 50.33%; FB1: 38.30 2704
PER: precision: 47.92%; recall: 52.13%; FB1: 49.94 1759
[Info] Save the RNN model in .//models//model_RNN_preTrue_biTrue_256_256_0.001.pth
[Info] --------------------------------------------------------------------------------
[Demo] --- RNNType: LSTM | HiddenNodes: 256 | Bi-Direction: False | Pre-Trained: True ---
[Para] BatchSize=256, LearningRate=0.0010, MaxEpoch=1000, PerEpoch=5.
[Epoch 005] loss: 0.706, train acc: 88.246%, valid acc: 86.465%.
[Epoch 010] loss: 0.376, train acc: 92.657%, valid acc: 89.677%.
[Epoch 015] loss: 0.238, train acc: 95.207%, valid acc: 91.906%.
[Epoch 020] loss: 0.148, train acc: 97.166%, valid acc: 91.634%.
[Epoch 025] loss: 0.105, train acc: 98.096%, valid acc: 92.346%.
[Eval] Testing accuracy: 90.044%.
processed 46435 tokens with 5648 phrases; found: 6413 phrases; correct: 3305.
accuracy: 63.63%; (non-O)
accuracy: 90.04%; precision: 51.54%; recall: 58.52%; FB1: 54.80
LOC: precision: 66.49%; recall: 68.05%; FB1: 67.26 1707
MISC: precision: 39.72%; recall: 48.15%; FB1: 43.53 851
ORG: precision: 42.39%; recall: 54.30%; FB1: 47.61 2128
PER: precision: 53.85%; recall: 57.51%; FB1: 55.62 1727
[Info] Save the LSTM model in .//models//model_LSTM_preTrue_biFalse_256_256_0.001.pth
[Info] --------------------------------------------------------------------------------
[Demo] --- RNNType: LSTM | HiddenNodes: 256 | Bi-Direction: True | Pre-Trained: True ---
[Para] BatchSize=256, LearningRate=0.0010, MaxEpoch=1000, PerEpoch=5.
[Epoch 005] loss: 0.519, train acc: 90.576%, valid acc: 88.014%.
[Epoch 010] loss: 0.216, train acc: 95.361%, valid acc: 91.067%.
[Epoch 015] loss: 0.112, train acc: 97.286%, valid acc: 93.201%.
[Epoch 020] loss: 0.042, train acc: 99.197%, valid acc: 93.112%.
[Epoch 025] loss: 0.021, train acc: 99.626%, valid acc: 93.421%.
[Eval] Testing accuracy: 91.545%.
processed 46435 tokens with 5648 phrases; found: 5587 phrases; correct: 3434.
accuracy: 64.36%; (non-O)
accuracy: 91.55%; precision: 61.46%; recall: 60.80%; FB1: 61.13
LOC: precision: 71.97%; recall: 72.96%; FB1: 72.46 1691
MISC: precision: 48.78%; recall: 51.42%; FB1: 50.07 740
ORG: precision: 54.36%; recall: 54.36%; FB1: 54.36 1661
PER: precision: 63.75%; recall: 58.94%; FB1: 61.25 1495
[Info] Save the LSTM model in .//models//model_LSTM_preTrue_biTrue_256_256_0.001.pth
[Info] --------------------------------------------------------------------------------
[Demo] --- RNNType: GRU | HiddenNodes: 256 | Bi-Direction: False | Pre-Trained: True ---
[Para] BatchSize=256, LearningRate=0.0010, MaxEpoch=1000, PerEpoch=5.
[Epoch 005] loss: 0.704, train acc: 88.091%, valid acc: 87.693%.
[Epoch 010] loss: 0.391, train acc: 92.372%, valid acc: 89.609%.
[Epoch 015] loss: 0.243, train acc: 95.143%, valid acc: 91.513%.
[Epoch 020] loss: 0.173, train acc: 96.613%, valid acc: 92.259%.
[Epoch 025] loss: 0.117, train acc: 97.708%, valid acc: 92.823%.
[Eval] Testing accuracy: 90.253%.
processed 46435 tokens with 5648 phrases; found: 6188 phrases; correct: 3289.
accuracy: 61.88%; (non-O)
accuracy: 90.25%; precision: 53.15%; recall: 58.23%; FB1: 55.58
LOC: precision: 62.53%; recall: 69.12%; FB1: 65.66 1844
MISC: precision: 48.19%; recall: 52.99%; FB1: 50.47 772
ORG: precision: 46.29%; recall: 49.91%; FB1: 48.03 1791
PER: precision: 52.50%; recall: 57.82%; FB1: 55.03 1781
[Info] Save the GRU model in .//models//model_GRU_preTrue_biFalse_256_256_0.001.pth
[Info] --------------------------------------------------------------------------------
[Demo] --- RNNType: GRU | HiddenNodes: 256 | Bi-Direction: True | Pre-Trained: True ---
[Para] BatchSize=256, LearningRate=0.0010, MaxEpoch=1000, PerEpoch=5.
[Epoch 005] loss: 0.531, train acc: 90.285%, valid acc: 88.998%.
[Epoch 010] loss: 0.251, train acc: 94.488%, valid acc: 92.309%.
[Epoch 015] loss: 0.118, train acc: 97.423%, valid acc: 92.751%.
[Epoch 020] loss: 0.069, train acc: 98.385%, valid acc: 92.467%.
[Eval] Testing accuracy: 91.089%.
processed 46435 tokens with 5648 phrases; found: 5971 phrases; correct: 3488.
accuracy: 65.95%; (non-O)
accuracy: 91.09%; precision: 58.42%; recall: 61.76%; FB1: 60.04
LOC: precision: 71.44%; recall: 70.62%; FB1: 71.03 1649
MISC: precision: 50.25%; recall: 56.70%; FB1: 53.28 792
ORG: precision: 49.41%; recall: 58.10%; FB1: 53.40 1953
PER: precision: 60.05%; recall: 58.57%; FB1: 59.30 1577
[Info] Save the GRU model in .//models//model_GRU_preTrue_biTrue_256_256_0.001.pth
[Info] --------------------------------------------------------------------------------
[Demo] --- RNNType: LSTM | HiddenNodes: 256 | Bi-Direction: True | Pre-Trained: False ---
[Para] BatchSize=256, LearningRate=0.0001, MaxEpoch=1000, PerEpoch=5.
[Epoch 005] loss: 1.281, train acc: 84.772%, valid acc: 85.293%.
[Epoch 010] loss: 0.830, train acc: 88.486%, valid acc: 88.334%.
[Epoch 015] loss: 0.567, train acc: 91.609%, valid acc: 90.148%.
[Epoch 020] loss: 0.406, train acc: 93.615%, valid acc: 91.003%.
[Epoch 025] loss: 0.307, train acc: 94.949%, valid acc: 92.278%.
[Epoch 030] loss: 0.229, train acc: 96.085%, valid acc: 92.839%.
[Epoch 035] loss: 0.173, train acc: 96.965%, valid acc: 92.925%.
[Eval] Testing accuracy: 90.557%.
processed 46435 tokens with 5648 phrases; found: 6805 phrases; correct: 3551.
accuracy: 69.43%; (non-O)
accuracy: 90.56%; precision: 52.18%; recall: 62.87%; FB1: 57.03
LOC: precision: 65.07%; recall: 76.74%; FB1: 70.43 1967
MISC: precision: 44.09%; recall: 53.70%; FB1: 48.43 855
ORG: precision: 42.84%; recall: 55.63%; FB1: 48.40 2157
PER: precision: 53.12%; recall: 59.99%; FB1: 56.35 1826
[Info] Save the LSTM model in .//models//model_LSTM_preFalse_biTrue_256_256_0.0001.pth
[Info] --------------------------------------------------------------------------------