11# -*- coding: UTF-8 -*-
22import numpy as np
3- #import jieba
43from seg_dnn import SegDNN
54from seg_lstm import SegLSTM
65from utils import estimate_cws
76import constant
87
98
109def test_seg_dnn ():
11- '''
1210 cws = SegDNN (constant .VOCAB_SIZE , 50 , constant .DNN_SKIP_WINDOW )
13- sentence = '迈向充满希望的新世纪'
14- # model = 'tmp/4.29-w3-normal/model15.ckpt'
15- model = 'tmp/4.29-100/model99.ckpt'
1611 model = 'tmp/model0.ckpt'
1712 print (cws .seg ('小明来自南京师范大学' , model ))
1813 print (cws .seg ('小明是上海理工大学的学生' , model ))
1914 print (cws .seg ('小明是清华大学的学生' , model ))
2015 print (cws .seg ('我爱北京天安门' , model ))
2116 print (cws .seg ('上海理工大学' , model ))
2217 print (cws .seg ('上海海洋大学' ))
23- print(cws.seg(sentence, model))
24- #print('/'.join(jieba.cut('小明是上海理工大学的学生')))
25- seq = cws.index2seq(cws.sentence2index(sentence))
26- seq = np.array(seq, dtype=np.int32).flatten()
27- '''
28- seg = SegLSTM ()
29- # seg.train_exe()
30- #print(seg.seg('我爱北京天安门'))
31- #print(seg.seg('小明来自南京师范大学'))
32- #print(seg.seg('小明是上海理工大学的学生'))
33- print (seg .seg ('小明来自南京师范大学' ))
34- print (seg .seg ('小明是上海理工大学的学生' ))
35- test (seg ,'tmp/lstm-model0.ckpt' )
36- # print(seq)
37- # cal_val(seq)
38- # print(cws.seg('2015世界旅游小姐大赛山东赛区冠军总决赛在威海举行',model))
18+ print (cws .seg ('迈向充满希望的新世纪' , model ))
3919
4020
41- def cal_val (seq ):
42- embeddings = np .load ('data/dnn/embeddings.npy' )
43- w2 = np .load ('data/dnn/w2.npy' )
44- w3 = np .load ('data/dnn/w3.npy' )
45- b2 = np .load ('data/dnn/b2.npy' )
46- b3 = np .load ('data/dnn/b3.npy' )
47- # b2 = np.expand_dims(b2.flatten(),0)
48- # A = np.load('data/dnn/A.npy')
49- # init_A = np.load('data/dnn/init_A.npy')
50- # print(w2)
51- # x = np.reshape(embeddings[seq],[10,150])
52- # print(x[0])
53- # b2 = np.tile(b2,[10])
54- # print(np.matmul(w2,x.T))
55- # print(b2.shape)
56- # val = np.matmul(w3,sigmoid(np.add(np.matmul(w2,x.T),b2)))+b3
57- # val = w3*sigmoid(w2*x.T+b2)+b3
58- # print(val.T)
21+ def test_seg_lstm ():
22+ seg = SegLSTM ()
23+ model = 'tmp/lstm-model1.ckpt'
24+ print (seg .seg ('小明来自南京师范大学' , model , debug = True ))
25+ print (seg .seg ('小明是上海理工大学的学生' , model ))
26+ print (seg .seg ('迈向充满希望的新世纪' , model ))
27+ print (seg .seg ('2015世界旅游小姐大赛山东赛区冠军总决赛在威海举行' , model ))
28+ test (seg , model )
5929
6030
6131def test (cws , model ):
@@ -66,31 +36,22 @@ def test(cws, model):
6636 corr_count = 0
6737 re_count = 0
6838 total_count = 0
39+ 6940 for _ , (sentence , label ) in enumerate (zip (sentences , labels )):
7041 label = np .array (list (map (lambda s : int (s ), label .split (' ' ))))
7142 _ , tag = cws .seg (sentence , model )
7243 cor_count , prec_count , recall_count = estimate_cws (tag , np .array (label ))
7344 corr_count += cor_count
7445 re_count += recall_count
7546 total_count += prec_count
76- # if(corr_count != prec_count):
77- # print(cws.tags2words(sentence,tag))
78- 79- # diff = np.subtract(tag,np.array(label))
80- # if sum()
81- # print(np.where(diff == 0))
82- # corr_count += len(np.where(diff == 0)[0])
83- # total_count += len(label)
8447 prec = corr_count / total_count
8548 recall = corr_count / re_count
49+ 8650 print (prec )
8751 print (recall )
8852 print (2 * prec * recall / (prec + recall ))
8953
9054
91- def sigmoid (x ):
92- return 1 / (1 + np .exp (- x ))
93- 94- 9555if __name__ == '__main__' :
96- test_seg_dnn ()
56+ # test_seg_dnn()
57+ test_seg_lstm ()
0 commit comments