-
Notifications
You must be signed in to change notification settings - Fork 1
/
tokenization.py
356 lines (300 loc) · 10.7 KB
/
tokenization.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
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
from __future__ import print_function, division
import os
from nltk.tag import StanfordPOSTagger
import xml.etree.ElementTree as ET
import networkx as nx
standford_pos_tagger = None
class Token:
def __init__(self, span, string, index, par=None):
self.span = span
self.string = string
self.next = None
self.previous = None
self.index = None
self.paragraph = par
self.entity_id = None
self.pos = 'NEWLINE' if string == '\n' else 'UNK_POS'
def __str__(self):
return str(self.string)
def set_next(self, next):
self.next = next
def set_previous(self, previous):
self.previous = previous
def set_index(self, index):
self.index = index
def get_string(self):
return str(self.string)
def get_index(self):
return self.index
class Tokenization:
def __init__(self):
self.tokens = []
self.num_tokens = 0
self.max_character_index = 0
self.start_index_to_token_index = {}
self.end_index_to_token_index = {}
self.paragrah_starts = None
self.dependencies = nx.DiGraph()
def __str__(self):
return str([str(token) for token in self.tokens])
def assign_tokens_to_entity(self, entity, token_assignment=True):
entity.tokens = self.tokens_in_span(entity.get_span())
if token_assignment:
for token in entity.tokens:
token.entity_id = entity.ID()
return entity
def same_paragraph(self,token1,token2):
return token1.paragraph == token2.paragraph
def assign_paragraph_to_entities(self,entities):
for e in entities:
e.paragraph = self.get_paragraph(e.get_span())
def get_paragraph(self, span):
for i,p in enumerate(self.paragrah_starts):
if span[0] < p:
return i
return len(self.paragrah_starts)
def closest_left_entity(self, entity, distance = True):
d = 0
if len(entity.tokens) > 0:
tok_index = max(entity.tokens[0].index,0)
else:
tok_index = self.first_left_from_span(entity.get_span()).index
for i in range(tok_index,0, -1):
d += 1
if self.tokens[i].entity_id and self.tokens[i].entity_id!=entity.ID():
if distance:
return self.tokens[i].entity_id, d
else:
return self.tokens[i].entity_id
if distance:
return None, d
else:
return None
def closest_right_entity(self, entity, distance = False):
d = 0
if len(entity.tokens) > 0:
tok_index = entity.tokens[-1].index
else:
tok_index = self.first_right_from_span(entity.get_span()).index
for i in range(tok_index, self.num_tokens):
d += 1
if self.tokens[i].entity_id and self.tokens[i].entity_id!=entity.ID():
if distance:
return self.tokens[i].entity_id, d
else:
return self.tokens[i].entity_id
if distance:
return None, d
else:
return None
def POS_tag(self, pos_model):
global standford_pos_tagger
if not standford_pos_tagger:
standford_pos_tagger = StanfordPOSTagger(os.getcwd() + '/' + pos_model,os.getcwd() + '/stanford-postagger.jar')
exceptions = set(['\n','\t'])
selected_tokens = [t for t in self.tokens if not t.get_string() in exceptions]
tags = standford_pos_tagger.tag([t.get_string() for t in selected_tokens])
for token, (string, tag) in zip(selected_tokens,tags):
token.pos = tag
print('POS:',len(tags),len(selected_tokens))
if len(tags)!=len(selected_tokens):
print('POS ERROR: number of tokens, and POS-tags does not match.')
exit()
def first_left_from_span(self, span):
for i in range(span[0],0,-1) :
if i in self.end_index_to_token_index:
return self.tokens[self.end_index_to_token_index[i]]
return Token((0,0),'<start_of_text>')
def first_right_from_span(self, span):
for i in range(span[1],self.max_character_index):
if i in self.start_index_to_token_index:
return self.tokens[self.start_index_to_token_index[i]]
return Token((0,0),'<end_of_text>')
def n_right_tokens(self, entity, n):
n_tokens = []
if entity.tokens == []:
tok = self.first_right_from_span(entity.get_span())
else:
tok = entity.tokens[0]
n_tokens = []
for i in range(1,n):
tok = self.tokens[tok.next] if tok.next != None else None
if tok:
n_tokens.append(tok)
else:
return n_tokens
return n_tokens
def n_left_tokens(self, entity, n):
n_tokens = []
if entity.tokens == []:
tok = self.first_left_from_span(entity.get_span())
else:
tok = entity.tokens[0]
n_tokens = []
for i in range(1,n):
tok = self.tokens[tok.previous] if tok.previous != None else None
if tok:
n_tokens.append(tok)
else:
return n_tokens
return n_tokens
def first_left_verb(self, entity):
token = self.first_left_from_span(entity.get_span())
d = 0
while(token and token.paragraph == entity.paragraph):
if token.pos[0] == 'V':
return token,d
token = self.tokens[token.previous] if token.previous != None else None
d+=1
return Token((0,0),'NOVERB',-1), 100000000
def first_right_verb(self, entity):
token = self.first_right_from_span(entity.get_span())
d = 0
while(token and token.paragraph == entity.paragraph):
if token.pos[0] == 'V':
return token,d
token = self.tokens[token.next] if token.next != None else None
d+=1
return Token((0,0),'NOVERB',-1), 100000000
def assign_tokens_to_entities(self, entities):
for entity in entities:
self.assign_tokens_to_entity(entity)
def tokens_in_span(self, span):
start, end = None, None
for i in range(span[0],span[1]):
if start==None and i in self.start_index_to_token_index:
start = i
if start == None:
return []
for i in range(span[1],span[0],-1):
if end==None and i in self.end_index_to_token_index:
end = i
if end == None:
return []
return self.tokens[self.start_index_to_token_index[start]:self.end_index_to_token_index[end]+1]
def tokens_inbetween(self, eLink):
if not eLink.get_e1().tokens or not eLink.get_e2().tokens:
if eLink.get_e1().get_span()[0] < eLink.get_e2().get_span()[0]:
return self.tokens_in_span((eLink.get_e1().get_span()[-1], eLink.get_e2().get_span()[0]))
else:
return reversed(self.tokens_in_span((eLink.get_e2().get_span()[-1],eLink.get_e2().get_span()[0])))
if eLink.get_e1().tokens[0].index < eLink.get_e2().tokens[0].index:
return self.tokens[eLink.get_e1().tokens[-1].index + 1:eLink.get_e2().tokens[0].index]
elif eLink.get_e1().tokens[0].index >= eLink.get_e2().tokens[0].index:
return self.tokens[eLink.get_e1().tokens[0].index + 1:eLink.get_e2().tokens[-1].index:-1]
def token_distance_between_entities(self, e1, e2):
if not e1.tokens or not e2.tokens:
return self.token_distance_between_spans(e1.get_span(), e2.get_span())
elif e1.tokens[0].index <= e2.tokens[0].index:
return e2.tokens[0].index - e1.tokens[-1].index
else:
return e2.tokens[-1].index - e1.tokens[0].index
def set_paragraph_starts(self, par_starts):
self.paragrah_starts = par_starts
def token_distance_between_spans(self, span1, span2):
if span1[0] < span2[0]:
s1,s2 = span1[-1],span2[0]
rev = 1
else:
s2,s1 = span1[0],span2[-1]+1
rev = -1
start, end = None, None
for i in range(s1,s2):
if start==None and i in self.start_index_to_token_index:
start = i
if start == None:
return 0
for i in range(s2,s1,-1):
if end==None and i in self.end_index_to_token_index:
end = i
if end == None:
return 0
return (self.end_index_to_token_index[end]+2 - self.start_index_to_token_index[start]) * rev
def append(self, token):
self.tokens.append(token)
self.start_index_to_token_index[token.span[0]] = self.num_tokens
self.end_index_to_token_index[token.span[1]] = self.num_tokens
token.set_index(self.num_tokens)
self.num_tokens += 1
self.max_character_index = token.span[1]
def read_ctakes(self, doc_id, ctakes_out_dir):
print('reading ctakes:',doc_id)
tree = ET.parse(ctakes_out_dir + doc_id + '.xml')
root = tree.getroot()
ctakes_POS = {} # from id to POS
ctakes_DEP = {} # from id to (id, label)
ctakes_span_to_id = {} # from span to id
span_to_index = {}
for tok in self.tokens:
span_to_index[tok.span] = tok.index
# reading xml file
for dep in root.iter('org.apache.ctakes.typesystem.type.syntax.ConllDependencyNode'):
if 'deprel' in dep.attrib:
label = dep.attrib['deprel']
head_id = int(dep.attrib['_ref_head'])
node_id = int(dep.attrib['_id'])
span = (int(dep.attrib['begin']), int(dep.attrib['end']))
pos = dep.attrib['postag']
ctakes_span_to_id[span] = node_id
ctakes_POS[node_id] = pos
ctakes_DEP[node_id] = (head_id, label)
# transforming ctakes ids to tokenization ids (token indices)
ctakes_id_to_token_id = {}
for span,id in ctakes_span_to_id.items():
token_index = None
if span in span_to_index:
token_index = span_to_index[span]
elif span[1] in self.end_index_to_token_index:
token_index = self.end_index_to_token_index[span[1]]
ctakes_id_to_token_id[id] = token_index
# assigning POS
for id,pos in ctakes_POS.items():
if ctakes_id_to_token_id[id]:
self.tokens[ctakes_id_to_token_id[id]].pos = pos
# assigning dependency relations
for id, (head_id, label) in ctakes_DEP.items():
if ctakes_id_to_token_id[id] and head_id in ctakes_id_to_token_id and ctakes_id_to_token_id[head_id]:
self.dependencies.add_edge(ctakes_id_to_token_id[id], ctakes_id_to_token_id[head_id], label=label + '>')
self.dependencies.add_edge(ctakes_id_to_token_id[head_id], ctakes_id_to_token_id[id], label='<' + label)
class SimpleTokenizer:
inclusive_splitters = set([',','.','/','\\','"','\n','=','+','-',';',':','(',')','!','?',"'",'<','>','%','&','$','*','|','[',']','{','}'])
exclusive_splitters = set([' ','\t'])
paragraph_splitters = set(['\n\n'])
def tokenize(self, text):
mem = ""
start = 0
paragraph_starts = [0]
par_mem = " "
tokens = Tokenization()
for i,char in enumerate(text):
par_mem = par_mem[-1] + char
if par_mem in self.paragraph_splitters:
paragraph_starts.append(i)
if char in self.inclusive_splitters:
if mem!="":
tokens.append(Token((start,i),text[start:i],len(tokens.tokens), par=len(paragraph_starts)))
mem = ""
tokens.append(Token((i,i+1),text[i:i+1],len(tokens.tokens), par=len(paragraph_starts)))
start = i+1
elif char in self.exclusive_splitters:
if mem!="":
tokens.append(Token((start,i),text[start:i],len(tokens.tokens), par=len(paragraph_starts)))
mem = ""
start = i+1
else:
start = i+1
else:
mem += char
tokens.set_paragraph_starts(paragraph_starts)
for i,t in enumerate(tokens.tokens):
if i+1 < len(tokens.tokens):
t.next = tokens.tokens[i+1].index
tokens.tokens[i+1].previous = t.index
return tokens
def test():
tokenizer = SimpleTokenizer()
text = "Over de vroege geschiedenis van Rome is veel geschreven."
tokenization = tokenizer.tokenize(text)
print([(t.span[0],t.span[1]) for t in tokenization.tokens])
print(tokenization)
print([str(t) for t in tokenization.tokens_in_span((0,15))])