lear_ner_extraction/soft_lexion_processing.py at main · qiufengyuyi/lear_ner_extraction · GitHub

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
174
175
176
# -*- coding: utf-8 -*-
import json
# import os
# import configparser
# # import pandas as pd
# import random
import re
import numpy as np
from gensim.models import word2vec, Word2Vec, keyedvectors, KeyedVectors, TfidfModel


def get_word_tf(embedding_file, training_context_datas):
    """
    simple lexicon 对出现在语料中词向量词表中的词，计算样本中每个字符分别在B,M,E,S中的向量表示
    先统计词频和出现的词表，（全量，包括测试集合训练集）
    :param embedding_file:
    :return:
    """
    vector_base = KeyedVectors.load_word2vec_format(embedding_file)
    # vector_base = KeyedVectors.load(embedding_file, mmap='r')
    embedding_vocab_list = vector_base.wv.index2word
    vocab_count_dict = {key: 0 for key in embedding_vocab_list}
    # {char+B:[word1,word2,...]}
    gaz_position_dict = {}
    text_list = []
    # for id,text_ele in training_context_datas.items():
    #     for key,text in text_ele.items():
    #         text_list.append(text)
    embedding_dict = {}
    text_str = "\n".join(training_context_datas)
    words_trim_list = []
    # words_ids_trim_list = []
    for index, vocab_local in enumerate(vocab_count_dict):
        if index % 500 == 0 and index > 0:
            print(index)
            # break
        try:
            match_count = len(re.findall(r"" + vocab_local, text_str))
            if match_count < 5:
                words_trim_list.append(vocab_local)
                continue
            vocab_count_dict[vocab_local] += len(re.findall(r"" + vocab_local, text_str))
            for index, char in enumerate(vocab_local):
                if len(vocab_local) == 1:
                    # single
                    if char in gaz_position_dict:
                        if "S" not in gaz_position_dict[char]:
                            gaz_position_dict[char]["S"] = [vocab_local]
                        else:
                            gaz_position_dict[char]["S"].append(vocab_local)
                    else:
                        gaz_position_dict[char] = {"S": [vocab_local]}
                        # gaz_position_dict[char]["S"] = [vocab_local]
                else:
                    if index == 0:

                        # start
                        if char in gaz_position_dict:
                            if "B" not in gaz_position_dict[char]:
                                gaz_position_dict[char]["B"] = [vocab_local]
                            else:
                                gaz_position_dict[char]["B"].append(vocab_local)
                        else:
                            gaz_position_dict[char] = {"B": [vocab_local]}
                    elif index < len(vocab_local) - 1:
                        # middle
                        if char in gaz_position_dict:
                            if "M" not in gaz_position_dict[char]:
                                gaz_position_dict[char]["M"] = [vocab_local]
                            else:
                                gaz_position_dict[char]["M"].append(vocab_local)
                        else:
                            gaz_position_dict[char] = {"M": [vocab_local]}
                    else:
                        # end
                        if char in gaz_position_dict:
                            if "E" not in gaz_position_dict[char]:
                                gaz_position_dict[char]["E"] = [vocab_local]
                            else:
                                gaz_position_dict[char]["E"].append(vocab_local)
                        else:
                            gaz_position_dict[char] = {"E": [vocab_local]}

            # if vocab_count_dict[vocab_local] < 5:
            #     # id_vocab = vector_base.wv.vocab[vocab_local].index
            #     # del vector_base.wv.vocab[vocab_local]
            #     # vector_base.wv.vectors =

            #     # words_ids_trim_list.append(vector_base.wv.vocab[vocab_local].index)
            #     # embedding_dict[vocab] = np.array(vector_base[vocab]).tolist()
        except Exception as e:
            print(e)
            continue
    # for id,text_ele in training_context_datas.items():
    #     for key,text in text_ele.items():
    #         for vocab in vocab_count_dict:
    #             try:
    #                 vocab_count_dict[vocab] += len(re.findall(r""+vocab,text))
    #             except:
    #                 continue
    # vocab_count_dict = {key:value for key,value in vocab_count_dict.items() if value > 0}
    # for w in words_trim_list:
    #     del vector_base.wv.vocab[w]
    # vector_base.wv.vectors = np.delete(vector_base.wv.vectors, words_ids_trim_list, axis=0)
    # vector_base.wv.init_sims(replace=True)
    # for i in sorted(words_ids_trim_list, reverse=True):
    #     del(vector_base.wv.index2word[i])
    # vector_base.save('data/bio_word2vec_trim')
    restrict_w2v(vector_base,words_trim_list)
    return vocab_count_dict, words_trim_list, gaz_position_dict


def get_text_from_raw_file(datas_file):
    text_list = []
    with open(datas_file, mode='r', encoding='utf-8') as fr:
        for index, line in enumerate(fr):
            line = line.strip("\n")
            if index % 2 == 0:
                char_list = line.split(" ")
                merge_text = "".join(char_list)
                text_list.append(merge_text)
    return text_list


def restrict_w2v(vector_base, restricted_word_set):
    """
    根据词表，将频率较高的词取出来，减小embedding的规模
    :param vector_base:
    :param restricted_word_set:
    :return:
    """
    new_vectors = []
    new_vocab = {}
    new_index2entity = []
    # new_vectors_norm = []

    for i in range(len(vector_base.vocab)):
        word = vector_base.index2entity[i]
        vec = vector_base.vectors[i]
        vocab = vector_base.vocab[word]
        # vec_norm = vector_base.vectors_norm[i]
        if word not in restricted_word_set:
            vocab.index = len(new_index2entity)
            new_index2entity.append(word)
            new_vocab[word] = vocab
            new_vectors.append(vec)
            # new_vectors_norm.append(vec_norm)

    vector_base.vocab = new_vocab
    vector_base.vectors = np.array(new_vectors)
    vector_base.index2entity = new_index2entity
    vector_base.index2word = new_index2entity
    # vector_base.vectors_norm = new_vectors_norm
    vector_base.init_sims(replace=True)
    # vector_base.save('bio_word2vec')
    vector_base.save('data/bio_word2vec_trim')


if __name__ == "__main__":
    # word2vec = KeyedVectors.load_word2vec_format("data/sgns.financial.bigram-char", binary=False,unicode_errors='ignore')
    # word2vec.init_sims(replace=True)

    embedding_file = "data/sgns.financial.bigram-char"
    training_context_datas = get_text_from_raw_file("data/orig_data_train.txt")
    training_context_datas.extend(get_text_from_raw_file("data/orig_data_dev.txt"))
    training_context_datas.extend(get_text_from_raw_file("data/orig_data_test.txt"))
    vocab_count_dict, words_trim_list, gaz_position_dict = get_word_tf(embedding_file, training_context_datas)
    # vocab_count_dict = {key: value for key, value in vocab_count_dict.items() if value > 0}

    # with open("data/soft_vocab_count.json",mode='r',encoding='utf-8') as fr:
    #     vocab_count_dict = json.load(fr)
    vocab_count_dict = {key: value for key, value in vocab_count_dict.items() if value >= 5}
    with open("data/soft_vocab_count.json", mode='w', encoding='utf-8') as fw1:
        json.dump(vocab_count_dict, fw1, ensure_ascii=False, indent=4)
    with open("data/gaz_position_dict.json", mode='w', encoding='utf-8') as fw2:
        json.dump(gaz_position_dict, fw2, ensure_ascii=False, indent=4)