-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathterm_matching_baseline.py
173 lines (154 loc) · 7.18 KB
/
term_matching_baseline.py
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
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
from sklearn.feature_extraction.text import TfidfTransformer
import numpy as np
from nltk.tokenize import word_tokenize
from nltk.stem.snowball import SnowballStemmer
import nltk
from gensim.summarization import bm25
nltk.download('punkt')
stemmer = SnowballStemmer("english")
def load_data(training_file, validation_file, testing_file):
# Load data from files
texts = []
labels = []
label_texts = []
for file_n in [training_file, validation_file, testing_file]:
txts = []
lbs = []
lb_txts = []
with open(file_n,'r',encoding='utf-8', errors='ignore') as f:
for line in f:
line = line.strip()
label, label_text, text = line.split("\t")
txts.append(text)
lbs.append(label)
lb_txts.append(label_text)
texts.append(txts)
labels.append(lbs)
label_texts.append(lb_txts)
return texts, label_texts, labels
def get_vocab(texts,label_texts,labels):
x = label_texts[0] #+ texts[1]
label_texts = label_texts[0] #+label_texts[1]
y = labels[0] # + labels[1]
label_y_dict = {}
label_x_dict = {}
for index, label_text in enumerate(label_texts):
if label_text not in label_y_dict:
label_y_dict[label_text] = y[index]
label_x_dict[label_text] = [x[index]]
else:
label_x_dict[label_text] +=[x[index]]
label_texts_new = list(label_y_dict.keys())
label_text_new_index = {index:label_text for index,label_text in enumerate(label_texts_new)}
vocab = list()
for label, x_texts in label_x_dict.items():
text_items = x_texts+[label]
for text_item in text_items:
for label_token in word_tokenize(text_item):
label_token_stem = stemmer.stem(label_token.lower())
if label_token_stem not in vocab:
vocab.append(label_token_stem)
vocab_dict = {token_stem: index for index, token_stem in enumerate(vocab)}
return vocab_dict, label_x_dict,label_y_dict, label_text_new_index,label_texts_new
def get_scores(tf_idf_tests,tf_idf_label,label_text_index,label_code_dict,y_test_text):
# print(np.max(np.dot(tf_idf_tests,tf_idf_label.T)))
labels_index = np.argmax(np.dot(tf_idf_tests,tf_idf_label.T),axis=1)
codes = [label_code_dict[label_text_index[label_index]] for label_index in labels_index]
#print(codes)
#print(y_test_text)
acc = 0.0
for index , code in enumerate(codes):
if code == y_test_text[index]:
acc+=1.0
score = acc / len(y_test_text)
return score
def get_bm_25_score(query_list, bm25_list, label_text_new_index, label_y_dict, y_test_text):
bm25Model = bm25.BM25(bm25_list)
average_idf = sum(map(lambda k: float(bm25Model.idf[k]), bm25Model.idf.keys())) / len(bm25Model.idf.keys())
codes = []
for query in query_list:
scores = bm25Model.get_scores(query, average_idf)
idx = scores.index(max(scores))
codes.append(label_y_dict[label_text_new_index[idx]])
acc = 0.0
for index , code in enumerate(codes):
if code == y_test_text[index]:
acc+=1.0
score = acc / len(y_test_text)
return score
def process(train_path, dev_path, test_path):
x_text, label_text, y_text = load_data(train_path,dev_path,test_path)
vocab_dict, label_x_dict, label_y_dict, label_text_new_index, label_texts_new = get_vocab(x_text,label_text,y_text)
text_matrix = np.zeros((len(label_texts_new)+len(x_text[1])+len(x_text[2]),len(vocab_dict)))
index = 0
bm25_list = []
query_dev_list = []
query_test_list = []
for label_text in label_texts_new:
texts = label_x_dict[label_text]
texts.append(label_text)
word_ids = np.zeros(len(vocab_dict))
for text in texts:
token_stem_list = []
for _, token in enumerate(word_tokenize(text)):
token_norm = stemmer.stem(token.lower())
word_ids[vocab_dict[token_norm]]+=1.0
token_stem_list.append(token_norm)
text_matrix[index] = word_ids
bm25_list.append(token_stem_list)
index+=1
for dev_text in x_text[1]:
word_ids = np.zeros(len(vocab_dict))
token_stem_list = []
for _, token in enumerate(word_tokenize(dev_text)):
token_norm = stemmer.stem(token.lower())
if token_norm in vocab_dict:
word_ids[vocab_dict[token_norm]] += 1.0
token_stem_list.append(token_norm)
text_matrix[index] = word_ids
query_dev_list.append(token_stem_list)
index += 1
for test_text in x_text[2]:
word_ids = np.zeros(len(vocab_dict))
token_stem_list = []
for _, token in enumerate(word_tokenize(test_text)):
token_norm = stemmer.stem(token.lower())
if token_norm in vocab_dict:
word_ids[vocab_dict[token_norm]] += 1.0
token_stem_list.append(token_norm)
text_matrix[index] = word_ids
query_test_list.append(token_stem_list)
index += 1
transformer = TfidfTransformer(smooth_idf=True,use_idf=True)
tfidf = transformer.fit_transform(text_matrix)
tf_idf_matrix = tfidf.toarray()
tf_idf_label = tf_idf_matrix[:len(label_texts_new)]
tf_idf_dev = tf_idf_matrix[len(label_texts_new):len(label_texts_new)+len(x_text[1])]
tf_idf_test = tf_idf_matrix[len(label_texts_new)+len(x_text[1]):]
return get_scores(tf_idf_dev,tf_idf_label,label_text_new_index,label_y_dict,y_text[1]),\
get_scores(tf_idf_test,tf_idf_label,label_text_new_index,label_y_dict,y_text[2]), \
get_bm_25_score(query_dev_list, bm25_list, label_text_new_index, label_y_dict, y_text[1]),\
get_bm_25_score(query_test_list,bm25_list, label_text_new_index, label_y_dict, y_text[2])
def term_matching_baseline(dataset):
scores_tf_idf_dev = 0.0
scores_bm25_dev = 0.0
scores_tf_idf_test = 0.0
scores_bm25_test = 0.0
print("Run Term Matching Baseline for %s: ..." %(dataset))
for i in range(10):
score_tf_idf_dev, score_tf_idf_test, score_bm25_dev, score_bm25_test = process("data/"+dataset + "/"+dataset+".fold-"+ str(i) +".train.txt",
"data/"+dataset + "/"+dataset+".fold-"+ str(i) +".validation.txt",
"data/"+dataset + "/"+dataset+".fold-"+ str(i) +".test.txt")
#print("Folder "+str(i) +" Dev accuracy: TF-IDF ", score_tf_idf_dev)
#print("Folder " + str(i) + " Dev accuracy: BM25 ", score_bm25_dev)
#print("Folder "+str(i) +" Test accuracy: TF-IDF ", score_tf_idf_test)
#print("Folder " + str(i) + " Test accuracy: BM25 ", score_bm25_test)
scores_tf_idf_dev+=score_tf_idf_dev
scores_bm25_dev +=score_bm25_dev
scores_tf_idf_test +=score_tf_idf_test
scores_bm25_test +=score_bm25_test
print("Average Dev accuracy for %s: TF-IDF %s " %(dataset, scores_tf_idf_dev/10.0))
print("Average Dev accuracy for %s: BM25 %s" %(dataset, scores_bm25_dev/10.0))
print("Average Test accuracy for %s: TF-IDF %s " %(dataset, scores_tf_idf_test/10.0))
print("Average Test accuracy for %s: BM25 %s" %(dataset, scores_bm25_test/10.0))
# term_matching_baseline("AskAPatient")