-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathSentencesPreprocessor.py
More file actions
173 lines (164 loc) · 5.98 KB
/
SentencesPreprocessor.py
File metadata and controls
173 lines (164 loc) · 5.98 KB
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 nltk.tree import Tree
from SentenceParser import SentenceParser
from treehelpers import *
class SentencesPreprocessor():
def __init__(self, sentences):
self.sentences = sentences
def generate_simple_sentences(self):
"""
Returns a list of preprocessed sentences that should be easier for making
questions.
"""
simple_sentences = []
for sentence in self.sentences:
# print sentence
try:
t = SentenceParser.parse(sentence)
except:
continue
orig_sentence = sentence_join(t)
self.__remove_node_type(t, 'PRN')
self.__remove_nickname_structures(t)
simples = self.__extract_simples(t, simple_sentences)
# for simple in simples:
# print sentence_join(simple)
# if (len(simples) == 0 or (len(simples) == 1 and sentence_join(simples[0]) == orig_sentence)):
# print orig_sentence
# print t
simple_sentences.extend(simples)
return simple_sentences
def __remove_node_type(self, t, type):
to_remove = []
for i in xrange(len(t)):
if (isinstance(t[i], Tree)):
if t[i].label() == type:
if i - 1 >= 0 and t[i - 1].label() == ',':
to_remove.append(i-1)
to_remove.append(i)
else:
to_remove.append(i)
else:
self.__remove_node_type(t[i], type)
if len(to_remove) == 1:
del t[to_remove[0]]
elif len(to_remove) == 2:
del t[to_remove[0]:to_remove[1]+1]
def __remove_comma_modifiers(self, t):
to_remove = []
for i in xrange(len(t)):
if (isinstance(t[i], Tree)):
comma_labels = ['SBAR', 'S', 'PP']
if t[i].label() in comma_labels:
if i - 1 >= 0 and t[i - 1].label() == ',':
to_remove.append(i-1)
to_remove.append(i)
else:
self.__remove_comma_modifiers(t[i])
if len(to_remove) == 1:
del t[to_remove[0]]
elif len(to_remove) == 2:
del t[to_remove[0]:to_remove[1]+1]
def __remove_nickname_structures(self, t):
if len(t) >= 3:
to_delete = -1
for i in xrange(len(t) - 2):
if (isinstance(t[i], Tree)):
if t[i].label() == "``" or t[i].label() == "''":
if t[i + 2].label() == "``" or t[i + 2].label() == "''":
to_delete = i
else:
self.__remove_nickname_structures(t[i])
if to_delete != -1:
del t[to_delete:to_delete+3]
else:
for i in xrange(len(t)):
if (isinstance(t[i], Tree)):
self.__remove_nickname_structures(t[i])
def __get_simples_from_np_vp(self, np, vp, pp=None):
simples = []
period = Tree.fromstring('(. .)')
if pp == None:
orig = Tree('S', children=[np, vp, period])
else:
orig = Tree('S', children=[np, vp, pp, period])
orig_no_pp = Tree('S', children=[np, vp, period])
# print sentence_join(orig)
# print sentence_join(vp)
# print immediate_labels(vp)
# print vp
if ['VP', 'CC', 'VP'] == immediate_labels(vp)[:3]:
simples.extend(self.__get_simples_from_np_vp(np, vp[0], pp))
simples.extend(self.__get_simples_from_np_vp(np, vp[2]))
elif ['VP', ',', 'CC', 'VP'] == immediate_labels(vp)[:4]:
simples.extend(self.__get_simples_from_np_vp(np, vp[0], pp))
simples.extend(self.__get_simples_from_np_vp(np, vp[3]))
else:
simples.append(orig)
if pp != None:
simples.append(orig_no_pp)
if len(simples) == 0:
simples.append(orig)
return simples
def __clean_nps(self, t, so_far):
if isinstance(t, str):
return
else:
if t.label() == 'NP':
no_replace = ['it', 'him', 'her', 'them']
if len(t) == 1 and t[0].label() == 'PRP' and t[0][0].lower() not in no_replace:
last = -1
while len(so_far) + last > -1 and len(so_far[last][0]) == 1 and so_far[last][0][0].label() == 'PRP':
last -= 1
if len(so_far) + last > -1:
t.extend(so_far[last][0])
del t[0]
if leftmost(t).label() not in ['NNP', 'NNPS']:
leftmost(t)[0] = leftmost(t)[0].lower()
if ',' in immediate_labels(t):
i = immediate_labels(t).index(',')
del t[i:]
if 'SBAR' in immediate_labels(t):
i = immediate_labels(t).index('SBAR')
if leftmost(t[i])[0] != 'to' and leftmost(t[i])[0] != 'that':
del t[i]
for s in t:
self.__clean_nps(s, so_far)
def __extract_simples(self, t, so_far):
simples = []
while len(t) == 1 and isinstance(t[0], Tree):
t = t[0]
self.__remove_comma_modifiers(t)
self.__clean_nps(t, so_far)
for s in t:
if isinstance(s, Tree) and s.label() == 'S':
simples.extend(self.__extract_simples(s, so_far))
for i in xrange(len(t) - 1):
if isinstance(t[i], Tree) and isinstance(t[i+1], Tree) and t[i].label() == 'NP' and t[i+1].label() == 'VP':
simples.extend(self.__get_simples_from_np_vp(t[i], t[i+1]))
skip_label_tests = False
for s in t:
if not isinstance(s, Tree):
skip_label_tests = True
break
if skip_label_tests:
return simples
if ['NP', 'VP', '.'] == immediate_labels(t):
simples.extend(self.__get_simples_from_np_vp(t[0], t[1]))
elif ['NP', 'ADVP', 'VP', '.'] == immediate_labels(t):
simples.extend(self.__get_simples_from_np_vp(t[0], t[2]))
elif ['S', ',', 'NP', 'VP', '.'] == immediate_labels(t):
del t[0:2]
simples.extend(self.__get_simples_from_np_vp(t[0], t[1]))
elif ['PP', ',', 'NP', 'VP', '.'] == immediate_labels(t):
leftmost(t)[0] = leftmost(t)[0].lower()
s = t[2]
s = leftmost(s)
s[0] = s[0][0].upper() + s[0][1:]
simples.extend(self.__get_simples_from_np_vp(t[2], t[3], t[0]))
elif ['SBAR', ',', 'NP', 'VP', '.'] == immediate_labels(t):
extender = Tree('PP', children=t[0])
leftmost(extender)[0] = leftmost(extender)[0].lower()
simples.extend(self.__get_simples_from_np_vp(t[2], t[3], extender))
# else:
# simples.append(t)
return simples