-
Notifications
You must be signed in to change notification settings - Fork 1
/
data.py
170 lines (128 loc) · 5.94 KB
/
data.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
from collections import defaultdict
import numpy as np
import random
import tokenization
class Example(object):
def __init__(self, doc_key, tokens, sentence_tokens, gold_starts, gold_ends, speaker_ids, cluster_ids, genre, document_index,
offset=0, bert_to_orig_map=None):
assert len(tokens) == len(speaker_ids)
self.doc_key = doc_key
self.tokens = tokens
self.sentence_tokens = sentence_tokens
self.gold_starts = gold_starts
self.gold_ends = gold_ends
self.speaker_ids = speaker_ids
self.cluster_ids = cluster_ids
self.genre = genre
self.document_index = document_index
self.offset = offset
self.bert_to_orig_map = bert_to_orig_map
def truncate(self, start, size):
# don't truncate in the middle of a mention
for mention in zip(self.gold_starts, self.gold_ends):
if index_in_mention(start, mention):
start = mention[0]
if index_in_mention(start + size, mention):
size -= start + size - mention[0]
end = start + size
tokens = self.tokens[start:end]
sentence_tokens = None
speaker_ids = self.speaker_ids[start:end]
gold_spans = np.logical_and(self.gold_starts >= start, self.gold_ends < end)
gold_starts = self.gold_starts[gold_spans] - start
gold_ends = self.gold_ends[gold_spans] - start
cluster_ids = self.cluster_ids[gold_spans]
return Example(self.doc_key, tokens, sentence_tokens, gold_starts, gold_ends, speaker_ids, cluster_ids,
self.genre, self.document_index, start)
def bertify(self, tokenizer):
assert self.offset == 0
bert_tokens = []
orig_to_bert_map = []
orig_to_bert_end_map = []
bert_speaker_ids = []
for t, s in zip(self.tokens, self.speaker_ids):
bert_t = tokenizer.tokenize(t)
orig_to_bert_map.append(len(bert_tokens))
orig_to_bert_end_map.append(len(bert_tokens) + len(bert_t) - 1)
bert_tokens.extend(bert_t)
bert_speaker_ids.extend([s] * len(bert_t))
bert_sentence_tokens = [tokenizer.tokenize(' '.join(s)) for s in self.sentence_tokens]
bert_to_orig_map = [-1] * len(bert_tokens)
for i, bert_i in enumerate(orig_to_bert_map):
bert_to_orig_map[bert_i] = i
orig_to_bert_map = np.array(orig_to_bert_map)
orig_to_bert_end_map = np.array(orig_to_bert_end_map)
if len(self.gold_starts):
gold_starts = orig_to_bert_map[self.gold_starts]
gold_ends = orig_to_bert_end_map[self.gold_ends]
else:
gold_starts = self.gold_starts
gold_ends = self.gold_ends
return Example(self.doc_key, bert_tokens, bert_sentence_tokens, gold_starts, gold_ends, bert_speaker_ids,
self.cluster_ids, self.genre, self.document_index, bert_to_orig_map=bert_to_orig_map)
def unravel_token_index(self, token_index):
prev_sentences_len = 0
for i, s in enumerate(self.sentence_tokens):
if token_index < prev_sentences_len + len(s):
token_index_in_sentence = token_index - prev_sentences_len
return i, token_index_in_sentence
prev_sentences_len += len(s)
raise ValueError('token_index is out of range ({} >= {})', token_index, len(self.tokens))
def index_in_mention(index, mention):
return mention[0] <= index and mention[1] >= index
def mention_contains(mention1, mention2):
return mention1[0] <= mention2[0] and mention1[1] >= mention2[1]
def filter_embedded_mentions(mentions):
"""
Filter out mentions embedded in other mentions
"""
filtered = []
for i, m in enumerate(mentions):
other_mentions = mentions[:i] + mentions[i + 1:]
if any(mention_contains(other_m, m) for other_m in other_mentions):
continue
filtered.append(m)
return filtered
def filter_overlapping_mentions(mentions):
start_to_mentions = defaultdict(list)
for m in mentions:
start_to_mentions[m[0]].append(m)
filtered_mentions = []
for ms in start_to_mentions.values():
if len(ms) > 1:
pass
max_mention = np.argmax([m[1] - m[0] for m in ms])
filtered_mentions.append(ms[max_mention])
return filtered_mentions
def flatten(l):
return [item for sublist in l for item in sublist]
def tensorize_mentions(mentions):
if len(mentions) > 0:
starts, ends = zip(*mentions)
else:
starts, ends = [], []
return np.array(starts), np.array(ends)
def process_example(example, index, genres, *, should_filter_embedded_mentions=False):
clusters = example["clusters"]
gold_mentions = sorted(tuple(m) for m in flatten(clusters))
if should_filter_embedded_mentions:
gold_mentions = filter_overlapping_mentions(gold_mentions)
# gold_mentions = filter_embedded_mentions(gold_mentions)
gold_mention_map = {m: i for i, m in enumerate(gold_mentions)}
cluster_ids = np.zeros(len(gold_mentions))
for cluster_id, cluster in enumerate(clusters):
for mention in cluster:
if tuple(mention) in gold_mention_map:
cluster_ids[gold_mention_map[tuple(mention)]] = cluster_id + 1
sentences = example["sentences"]
num_words = sum(len(s) for s in sentences)
speakers = flatten(example["speakers"])
assert num_words == len(speakers)
sentence_tokens = [[tokenization.convert_to_unicode(w) for w in s] for s in sentences]
tokens = sum(sentence_tokens, [])
speaker_dict = {s: i for i, s in enumerate(set(speakers))}
speaker_ids = np.array([speaker_dict[s] for s in speakers])
doc_key = example["doc_key"]
genre = genres[doc_key[:2]]
gold_starts, gold_ends = tensorize_mentions(sorted(gold_mentions))
return Example(doc_key, tokens, sentence_tokens, gold_starts, gold_ends, speaker_ids, cluster_ids, genre, index)