-
Notifications
You must be signed in to change notification settings - Fork 0
/
corefextraction.py
156 lines (136 loc) · 4.3 KB
/
corefextraction.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
# -*- coding: utf-8 -*-
import re
import torch
import spacy
import neuralcoref
from allennlp.predictors.predictor import Predictor
from nltk.tokenize import sent_tokenize, word_tokenize
class Tripple:
def __init__(self, arg0,verb,arg1):
self.subject = arg0.lower()
self.verb = verb.lower()
self.object = arg1.lower()
def __str__(self):
return f'{self.subject.capitalize()} {self.verb} {self.object}. '
def __repr__(self):
return f'Tripple({self.subject!r},{self.verb!r},{self.object!r})'
def __eq__(self, other):
return ((self.verb in other.verb) and (other.object in self.object)) or ((other.verb in self.verb) and (self.object in other.object))
def __len__(self):
return len(str(self))
class InformationExtractor():
def __init__(self, coreference = False):
self.predictor = Predictor.from_path("https://s3-us-west-2.amazonaws.com/allennlp/models/openie-model.2018-08-20.tar.gz")
if torch.cuda.is_available():
self.predictor._model = self.predictor._model.cuda(0)
self.spacy_pipeline = spacy.load('en')
self.coreference = coreference
if self.coreference:
coref = neuralcoref.NeuralCoref(self.spacy_pipeline.vocab)
self.spacy_pipeline.add_pipe(coref, name='neuralcoref')
def Arguments(self):
_dict = dict({})
def dict_instance(string):
values = string.split(': ')
if len(values) > 1:
_dict[values[0]] = values[1]
return _dict
return dict_instance
def find_tripples(self,string):
tripples = []
extraction = self.predictor.predict(
sentence=string
)
#print(extraction)
for phrase in extraction['verbs']:
args = dict({})
subject = None
action = None
object1 = None
object2 = None
matches=re.findall(r'\[(.+?)\]',phrase['description'])
for x in matches:
keyValues = x.split(': ')
if len(keyValues) > 1:
args[keyValues[0]] = keyValues[1]
if 'ARG0' in args:
subject = args['ARG0']
if 'ARG1' in args:
object1 = args['ARG1']
if 'ARG2' in args:
if object1 is not None:
object1 = object1 + ' ' + args['ARG2']
else:
object1 = args['ARG2']
if 'V' in args:
action = args['V']
if 'BV' in args:
action = args['BV'] +' '+action
if 'AV' in args:
action = action + ' ' + args['AV']
if subject and action and object1:
new_tripple = Tripple(subject,action,object1)
#tripples.append(new_tripple)
#print(new_tripple)
if len(tripples):
old_tripple = tripples[-1]
if old_tripple == new_tripple:
if len(new_tripple.verb) > len(old_tripple.verb):
tripples[-1] = new_tripple
else:
tripples.append(new_tripple)
else:
tripples.append(new_tripple)
#if tripples:
# return max(tripples, key=len)
#else:
# return None
return tripples
def process(self,text):
sentnces = self.sent_tokenize(text)
tripples = [self.find_tripples(sent) for sent in sentnces]
tripples =[sent for sent in tripples if sent is not None]
output = []
for tripple in tripples:
output += tripple
return output
def sent_tokenize(self,input_):
if not self.coreference:
if isinstance(input_,list):
sentences = input_
else:
document = self.spacy_pipeline(input_)
sentences = [str(sent) for sent in document.sents]
else:
if isinstance(input_,list):
document = self.spacy_pipeline(" ".join(input_))
sentences = input_
else:
document = self.spacy_pipeline(input_)
sentences = [str(sent) for sent in document.sents]
if document._.has_coref:
sentences = self.get_resolved(document, sentences)
output = sentences
return output
def get_resolved(self, doc, sentences):
def get_2d_element(arrays, index):
j = index
lens = [len(sent) for sent in arrays]
for i,length in enumerate(lens):
j = j - length
if j < 0:
return i, length + j
resolved_list = []
tokenizer = spacy.load('en')
for sent in sentences:
resolved_list.append(list(tok.text_with_ws for tok in tokenizer(sent)))
for cluster in doc._.coref_clusters:
for coref in cluster:
if coref != cluster.main:
ind1, ind2 = get_2d_element(resolved_list,coref.start)
resolved_list[ind1][ind2] = cluster.main.text + doc[coref.end-1].whitespace_
for i in range(coref.start+1, coref.end):
ind3, ind4 = get_2d_element(resolved_list,i)
resolved_list[ind3][ind4] = ""
output = [''.join(sublist) for sublist in resolved_list]
return output