NuPyTools/seq2seq_g2p_p.py at master · JeffpanUK/NuPyTools · 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
#!/usr/bin/env py3
# -*- coding:utf-8 -*-

import tensorflow as tf
from tensorflow.contrib import keras
from tensorflow.contrib.keras import layers
from tensorflow.contrib.keras import wrappers
from tensorflow.contrib.keras import models
from tensorflow.contrib.keras import utils
import os,sys
import re
import codecs
import numpy as np
import pandas as pd

class Seq2Seq_G2P_P(object):
  '''
  '''
  def __init__(self):
    print("initializing word segmentation...")

  def viterbi(self, nodes):
    transit_maxtrix = {'be':0.5,
                      'bm':0.5,
                      'eb':0.5,
                      'es':0.5,
                      'me':0.5,
                      'mm':0.5,
                      'sb':0.5,
                      'ss':0.5}
    transit_matrix = {i:np.log(transit_matrix[i]) for i in transit_matrix.keys()}
    paths = {'b':nodes[0]['b'], 's':nodes[0]['s']}
    for l in range(1,len(nodes)):
        paths_ = paths.copy()
        paths = {}
        for i in nodes[l].keys():
            nows = {}
            for j in paths_.keys():
                if j[-1]+i in transit_matrix.keys():
                    nows[j+i]= paths_[j]+nodes[l][i]+transit_matrix[j[-1]+i]
            k = np.argmax(nows.values())
            paths[nows.keys()[k]] = nows.values()[k]
    return paths.keys()[np.argmax(paths.values())]

  def simple_cut(self, s):
    if s:
      r = models.models.predict(np.array([list(chars[list(s)].fillna(0).astype(int)) \
                                  + [0]*(time_step-len(s))]),
                                  verbose=False)[0][:len(s)]
      r = np.log(r)
      nodes = [dict(zip(['s','b','m','e'], i[:4])) for i in r]
      t = self.viterbi(nodes)
      words = []
      for i in range(len(s)):
        if t[i] in ['s', 'b']:
          words.append(s[i])
        else:
          words[-1] += s[i]
      return words
    else:
        return []

  def cut_word(self, s):
    result = []
    j = 0
    for i in not_cuts.finditer(s):
        result.extend(self.simple_cut(s[j:i.start()]))
        result.append(s[i.start():i.end()])
        j = i.end()
    result.extend(self.simple_cut(s[j:]))
    return result

  def main(self, fn):
    data = []
    label = []
    tmp_data = []
    tmp_label = []
    train_data = codecs.open(fn, 'r', 'utf-8')
    for line in train_data:
      if len(line.strip().split()) == 0:
        data.append(tmp_data)
        label.append(tmp_label)
        tmp_data = []
        tmp_label = []
        continue
      line = line.strip().split()
      # print(line)
      tmp_data.append(line[0])
      tmp_label.append(line[-2])


    d = pd.DataFrame(index=range(len(data)))
    d['data'] = data
    d['label'] = label
    d.index = range(len(d))
    tag = pd.Series({'S':0, 'B':1, 'M':2, 'E':3})
    tagL = {'S':0, 'B':1, 'M':2, 'E':3}
    chars = [] #count all words and indexing them
    for i in data:
        chars.extend(i)

    chars = pd.Series(chars).value_counts()
    chars[:] = range(1, len(chars)+1)

    embedding_size = 200
    time_step = 50
    # print(tag)
    # print(d['label'])
    print(d['data'][0])
    print(chars[d['data'][0]])
    d['x'] = d['data'].apply(lambda x: np.array(list(chars[x])+[0]*(time_step-len(x))))
    print(len(d['label'][0]))
    input("pause...")
    # d['y'] = d['label'].apply(lambda x: np.array(utils.to_categorical(tag.reshape(-1,1),4)[tagL[x]]))
    d['y'] = d['label'].apply(lambda x: np.array(list(map(lambda y:utils.to_categorical(y,4), tag[x].reshape((-1,1))))))
    # d['y'] = d['label'].apply(lambda x: np.array(utils.to_categorical(tag.reshape(-1,1),4))+[np.array([[0,0,0,0,1]])]*(time_step-len(x)))


    sequence = layers.Input(shape=(time_step,),
                      dtype='int32')

    embedded = layers.Embedding(len(chars)+1,
                        embedding_size,
                        input_length=time_step,
                        mask_zero=True)(sequence)

    blayers.LSTM1 = wrappers.Bidirectional(layers.LSTM(256, return_sequences=True),
                          merge_mode='sum')(embedded)
    blayers.LSTM2 = wrappers.Bidirectional(layers.LSTM(256, return_sequences=True),
                          merge_mode='sum')(blayers.LSTM1)
    blayers.LSTM3 = wrappers.Bidirectional(layers.LSTM(256, return_sequences=True),
                          merge_mode='sum')(blayers.LSTM2)
    # blayers.LSTM3 = layers.Bidirectional(layers.LSTM(256, return_sequences=True),
                          # merge_mode='sum')(blayers.LSTM2)

    output = wrappers.TimeDistributed(layers.Dense(4, activation='softmax'))(blayers.LSTM3)

    models.models = models.models(input=sequence, output=output)
    models.models.compile(loss='categorical_crossentropy',
                  optimizer='adam',
                  metrics=['accuracy'])

    batch_size = 1024

    history = models.models.fit(np.array(list(d['x'])),
                        np.array(list(d['y'])).reshape((-1,time_step,)),
                        batch_size=batch_size,
                        nb_epoch=50)

if __name__ == '__main__':
  fn = r"D:\Documents\GitHub\Tools\data\corpus\ws.wapiti"
  ws = Seq2Seq_G2P_P()
  ws.main(fn)
  s = "今天天气不错"
  ws.cut(s)