NamedEntityTagger/5_2.py at master · AMaini503/NamedEntityTagger · 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
#!/usr/bin/python3
import numpy as np
import sys
from aux import computeEmissions, isRare, getTrigramCount, getBigramCount, getAllWordsAndTags

#return count(x, y) / count(y)
def getEmission(emissions, x, y):
    if(isRare(emissions, x)):
        return emissions['_RARE_'][y]
    else:
        return emissions[x][y]


def getq(q, u, v, w):
    if((u, v, w) in q):
        return q[(u, v, w)]
    else:
        return -1


#x is 0-indexed
def getT(x, k, emissions):
    if(k < 0):
        return ['*']
    else:
        if(isRare(emissions, x[k])):
            return list(emissions['_RARE_'].keys())
        else:
            return list(emissions[x[k]].keys())


def tagUsingViterbi(x, emissions, q):


    #initialization
    pi = dict()
    bp = dict()

    pi[(-1, '*', '*')] = 1

    #tagged sequence has same length as the input sequence
    y = x[:]

    for k in range(len(x)):
        for u in  getT(x, k - 1, emissions):
            for v in getT(x, k, emissions):
                pi[(k, u, v)] = 0
                bp[(k, u, v)] = ''

                for w in getT(x, k - 2, emissions):
                        q_ml = getq(q, w, u, v)
                        if(q_ml != -1):
                            this_probability = pi[(k - 1, w, u)] * q_ml * getEmission(emissions, x[k], v)
                            if(pi[(k, u, v)]  < this_probability):
                                pi[(k, u, v)] = this_probability
                                bp[(k, u, v)] = w

    max_prob = -1
    n = len(x)
    for u in getT(x, n - 2, emissions):
        for v in getT(x, n - 1, emissions):
            q_ml = getq(q, u, v, 'STOP')
            if(q_ml != -1):
                if(max_prob < pi[(n - 1, u, v)] * q_ml):
                    max_prob = pi[(n - 1, u, v)] * q_ml
                    y[n - 2] = u
                    y[n - 1] = v

    for k in range(n - 3, -1, -1):
        y[k] = bp[(k + 2, y[k + 1], y[k + 2])]

    return (y, pi)


#this function assumes that y is 1-indexed
def gety(y, k):
    if(k < 0):
        return '*'
    else:
        return y[k]

def getLogLikelihood(x, y, pi):
    log_likelihood = []
    for k in range(len(x)):
        log_prob = np.log(pi[(k, gety(y, k - 1), gety(y, k))])
        log_likelihood.append(log_prob)

    return log_likelihood


def writeThisSentence(f_output, x, y, log_likelihood):
    for (word, tag, log_prob) in zip(x, y, log_likelihood):
        this_line = ' '.join((word, tag, '{}'.format(log_prob)))
        this_line = this_line + '\n'
        f_output.write(this_line)
    f_output.write('\n')


def computeq():
    count_of_trigrams = dict()
    count_of_bigrams = dict()

    with open('ner_rare.counts') as f_input:
        for line in f_input:
            tokens = line.strip().split()
            if(tokens[1] == '2-GRAM'):
                u, v = tokens[2], tokens[3]
                count_of_bigrams[(u, v)] = int(tokens[0])
            elif(tokens[1] == '3-GRAM'):
                u, v, w = tokens[2], tokens[3], tokens[4]
                count_of_trigrams[(u, v, w)] = int(tokens[0])
    q = dict()
    for (u, v, w) in count_of_trigrams:
        q[(u, v, w)] = float(count_of_trigrams[(u, v, w)])/ float(count_of_bigrams[(u, v)])


    return q


#initial parameters
emissions = computeEmissions()
q = computeq()


with open('ner_dev.dat') as f_input, open('5_2.txt', 'w') as f_output:
    x = []
    for line in f_input:
        tokens = line.strip().split()

        if(len(tokens) > 0):
            x.append(tokens[0])
        else:
            #x has buffered a sentence
            y, pi = tagUsingViterbi(x, emissions, q)
            log_likelihood = getLogLikelihood(x, y, pi)
            writeThisSentence(f_output, x, y, log_likelihood)

            x = []