-
Notifications
You must be signed in to change notification settings - Fork 6
/
lsi_wiki_model.py
137 lines (106 loc) · 5.42 KB
/
lsi_wiki_model.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
from gensim.utils import ClippedCorpus
from gensim.models import LsiModel, TfidfModel
from gensim.corpora import Dictionary, MmCorpus
from datasets.topical_dataset import TopicalDataset
from datasets.wikidata import WikiData, WikiCorpus
from datasets.wiki_database import WikiDatabase
import numpy as np
import os
import pickle
from configs import LSIWikiConfig
from topical_tokenizers import SpacyTokenizer, TransformerGPT2Tokenizer
class LSIModelWiki:
def __init__(self, config_file):
self.config = LSIWikiConfig.from_json_file(config_file)
if self.config.tokenizer == "gpt":
self.tokenizer = TransformerGPT2Tokenizer(self.config.cached_dir)
self.wiki_dict_file = os.path.join(self.config.cached_dir, "wiki_dict")
self.mm_corpus_file = os.path.join(self.config.cached_dir, "wiki_bow.mm")
self.wiki_tfidf_file = os.path.join(self.config.cached_dir, "wiki_tfidf.mm")
self.wiki_lsi_file = os.path.join(self.config.cached_dir, "wiki_lsi.mm")
self.model_file = os.path.join(self.config.cached_dir, "lsi_model.p")
self.topic_words_matrix_file = os.path.join(self.config.cached_dir, "topic_words_matrix.p")
self.topic_top_words_file = os.path.join(self.config.cached_dir, "topic_top_words.p")
def _create_files_db(self):
config_file = "/home/rohola/codes/topical_language_generation/configs/wiki_database.json"
wikidata = WikiDatabase(config_file, self.tokenizer)
docs = []
for i, tokens in enumerate(wikidata):
docs.append(tokens)
if i > 50000:
break
id2word_wiki = Dictionary(docs)
id2word_wiki.filter_extremes(no_below=20, no_above=0.01)
id2word_wiki.save(self.wiki_dict_file)
wiki_corpus = WikiCorpus(docs, id2word_wiki)
MmCorpus.serialize(self.mm_corpus_file, wiki_corpus)
def _create_files1(self):
dir = "/media/rohola/data/dialog_systems/alexa_prize_topical_chat_dataset/reading_sets/"
wikidata = TopicalDataset(dir, self.tokenizer)
doc_stream = (tokens for tokens in wikidata)
id2word_wiki = Dictionary(doc_stream)
id2word_wiki.filter_extremes(no_below=20, no_above=0.2)
id2word_wiki.save(self.wiki_dict_file)
wiki_corpus = WikiCorpus(wikidata, id2word_wiki)
MmCorpus.serialize(self.mm_corpus_file, wiki_corpus)
def _create_files(self):
wikidata = WikiData(self.config.dataset_dir, self.tokenizer)
doc_stream = (tokens for tokens in wikidata)
id2word_wiki = Dictionary(doc_stream)
id2word_wiki.filter_extremes(no_below=20, no_above=0.1)
id2word_wiki.save(self.wiki_dict_file)
wiki_corpus = WikiCorpus(wikidata, id2word_wiki)
MmCorpus.serialize(self.mm_corpus_file, wiki_corpus)
def _run_model(self):
id2word_wiki = Dictionary.load(self.wiki_dict_file)
mm_corpus = MmCorpus(self.mm_corpus_file)
#to be removed
#mm_corpus = ClippedCorpus(mm_corpus, 4000)
tfidf_model = TfidfModel(mm_corpus, id2word=id2word_wiki)
corpus = tfidf_model[mm_corpus]
MmCorpus.serialize(self.wiki_tfidf_file, corpus)
self.model = LsiModel(corpus,
num_topics=self.config.num_topics,
id2word=id2word_wiki,
chunksize=self.config.chunksize)
MmCorpus.serialize(self.wiki_lsi_file, self.model[corpus])
self.model.save(self.model_file)
def get_model(self):
return LsiModel.load(self.model_file)
def get_topic_words_matrix(self):
if not os.path.isfile(self.topic_words_matrix_file):
lsi_model = self.get_model()
topic_words = lsi_model.get_topics() # K X V' (num_topics x selected_vocab_size)
topic_word_matrix = np.zeros(
(self.config.num_topics, self.tokenizer.tokenizer.vocab_size)) # K x V (num_topics x vocab_size)
for i in range(len(lsi_model.id2word)):
j = self.tokenizer.tokenizer.convert_tokens_to_ids(lsi_model.id2word[i])
topic_word_matrix[:, j] = topic_words[:, i]
pickle.dump(topic_word_matrix, open(self.topic_words_matrix_file, 'wb'))
else:
topic_word_matrix = pickle.load(open(self.topic_words_matrix_file, 'rb'))
return topic_word_matrix
def get_topic_words(self, num_words=None):
if not os.path.isfile(self.topic_top_words_file):
if not num_words:
num_words = len(self.dictionary)
model = self.get_model()
topic_words = model.show_topics(self.config.num_topics,
num_words=num_words,
formatted=False)
pickle.dump(topic_words, open(self.topic_top_words_file, 'wb'))
else:
topic_words = pickle.load(open(self.topic_top_words_file, 'rb'))
return topic_words
if __name__ == "__main__":
config_file = "/home/rohola/codes/topical_language_generation/configs/wiki_lsi_config.json"
lsi_model_wiki = LSIModelWiki(config_file)
lsi_model_wiki._create_files_db()
#lsi_model_wiki._create_files1()
lsi_model_wiki._run_model()
#m = lsi_model_wiki.get_model()
#tw = lsi_model_wiki.get_topic_words_matrix()
twords = lsi_model_wiki.get_topic_words(10)
topic_words =[t[1] for t in twords]
for topic in topic_words:
print(topic)