forked from spro/pytorch-seq2seq-intent-parsing
-
Notifications
You must be signed in to change notification settings - Fork 0
/
data.py
280 lines (235 loc) · 8.89 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
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
import torch
from torch.autograd import Variable
import random
import re
# ## Generating training data
templates = [
("lights setState $light_name $light_state", [
"~turn $light_state $light_name",
"~turn $light_name $light_state",
]),
("lights setState $light_name $light_amount", [
"~turn $light_name $light_amount",
]),
("lights setStates $group_name $light_state", [
"~turn $light_state $group_name",
"~turn $group_name $light_state",
]),
("lights setStates $group_name $light_amount", [
"~turn $group_name $light_amount",
]),
("lights getState $light_name", [
"is $light_name on",
]),
("lights getStates $group_name", [
"are $group_name on",
]),
("music setVolume $volume", [
"~turn the music $volume",
"~turn it $volume",
]),
("time getTime", [
"time",
"what time is it",
"~whatis the time",
]),
("price getPrice $asset", [
"$asset",
"how much is $asset",
"~price of $asset",
"~whatis the ~price of $asset",
]),
("weather getWeather $location", [
"tell me the weather in $location",
"~whatis it like in $location",
]),
("greeting", [
"hi", "hello", "how are you", "what's up", "hey maia",
]),
("thanks", [
"thanks", "thank you", "thank you so much", "thx", "you're great",
]),
]
variables = {
"$light_name": [
("office_light", ["the office light", "the light in the office"]),
("kitchen_light", ["the kitchen light", "the light in the kitchen"]),
("living_room_light", ["the living room light", "the light in the living room", "the light in the den"]),
("outside_light", ["the outside light", "the outdoor light", "the light outside", "the porch light"]),
],
"$group_name": [
("all_lights", ["all the lights", "everything"]),
("bedroom_lights", ["my lights", "the bedroom lights", "the lights", "the lights in my room"]),
],
"$light_state": [
("on", ["on"]),
("off", ["off", "out"]),
("blue", ["blue"]),
("green", ["green"]),
("red", ["red"]),
("purple", ["purple"]),
("orange", ["orange"]),
("white", ["white", "normal"]),
],
"$light_amount": [
("low", ["low", "dim", "dark"]),
("high", ["high", "bright"]),
("down", ["down", "lower", "darker"]),
("up", ["up", "brighter"]),
],
"$volume": [
("high", ["high", "loud"]),
("low", ["low", "quiet", "down"]),
("up", ["up", "louder"]),
("down", ["down", "quieter"]),
],
"$location": [
("san_francisco", ["sf", "san francisco", "the city"]),
("new_hampshire", ["nh", "new hampshire", "the northeast"]),
],
"$asset": [
("btc", ["btc", "bitcoin", "bitcoins"]),
("eth", ["eth", "ethereum", "etherium", "ether"]),
("usd", ["usd", "dollars", "us dollars", "the fed"]),
("pesos", ["pesos", "mexican dollars"]),
],
}
synonyms = {
"~turn": ["turn", "set", "make", "put", "change"],
"~whatis": ["what is", "what's", "whats", "tell me", "tell us", "tell me about", "what about", "how about", "show me"],
"~price": ["price", "value", "exchange rate", "dollar amount"],
}
prefixes = ["please", "pls", "hey maia", "hi", "could you", "would you", "hey", "yo", "ey", "oy", "excuse me please"]
suffixes = ["thanks", "thank you", "please", "plz", "pls", "plox", "ok", "thank you so much"]
# Choose a random pair of templates (output, input)
def choose_templates():
output_template, input_templates = random.choice(templates)
input_template = random.choice(input_templates)
return output_template, input_template
input_template, output_template = choose_templates()
print('input template =', input_template)
print('output template =', output_template)
# We'll assume that all the variables in the input template are used in the output template.
# Choose variable values to fill a template with (output, input)
def choose_variables(template):
variable_names = [word for word in template.split(' ') if word[0] == '$']
input_variables = {}
output_variables = {}
for variable_name in variable_names:
variable = random.choice(variables[variable_name])
output_variables[variable_name] = variable[0]
input_variables[variable_name] = random.choice(variable[1])
return output_variables, input_variables
input_variables, output_variables = choose_variables(input_template)
print('input variables =', input_variables)
print('output variables =', output_variables)
def fill_template(template, template_variables):
filled = []
for word in template.split(' '):
# Choose variable
if word[0] == '$':
filled.append(template_variables[word])
# Choose synonym
elif word[0] == '~':
filled.append(random.choice(synonyms[word]))
# Regular word
else:
filled.append(word)
return ' '.join(filled)
PREFIX_PROB = 0.3
SUFFIX_PROB = 0.3
def add_fixes(sentence):
if random.random() < PREFIX_PROB:
sentence = random.choice(prefixes) + ' ' + sentence
if random.random() < SUFFIX_PROB:
sentence += ' ' + random.choice(suffixes)
return sentence
def random_training_pair():
output_template, input_template = choose_templates()
output_variables, input_variables = choose_variables(input_template)
output_string = fill_template(output_template, output_variables)
input_string = fill_template(input_template, input_variables)
input_string = add_fixes(input_string)
return input_string, output_string
for i in range(10):
print('\n', random_training_pair())
# ## Keeping track of the output vocabulary
SOS_token = 0
EOS_token = 1
def tokenize_sentence(s):
s = re.sub('[^\w\s]', '', s)
s = re.sub('\s+', ' ', s)
return s.strip().split(' ')
class DictionaryLang():
def __init__(self):
self.word2index = {}
self.word2count = {}
self.index2word = {0: "SOS", 1: "EOS"}
self.size = 2 # Count SOS and EOS
def __str__(self):
return "%s(size = %d)" % (self.__class__.__name__, self.size)
def add_word(self, word):
if word not in self.word2index:
self.word2index[word] = self.size
self.word2count[word] = 1
self.index2word[self.size] = word
self.size += 1
else:
self.word2count[word] += 1
def get_word(self, word):
return self.word2index[word]
def indexes_from_sentence(self, sentence):
return [self.get_word(word) for word in tokenize_sentence(sentence)]
def variable_from_sentence(self, sentence):
indexes = self.indexes_from_sentence(sentence)
indexes.append(EOS_token)
return Variable(torch.LongTensor(indexes).view(-1, 1))
# First turn the generated data into Lang for input (english) and output (command) languages
output_lang = DictionaryLang()
def add_words(lang, template):
for word in template.split(' '):
if word[0] != '$':
lang.add_word(word)
# Add words from templates
for output_template, input_templates in templates:
add_words(output_lang, output_template)
# Add values of variables
for variable_name in variables:
for output_variable, input_variables in variables[variable_name]:
add_words(output_lang, output_variable)
print("output lang = %s" % output_lang)
# ## Using word vectors for the input vocabulary
from torchtext.vocab import load_word_vectors
class GloVeLang:
def __init__(self, size):
self.size = size
glove_dict, glove_arr, glove_size = load_word_vectors('data/', 'glove.twitter.27B', size)
self.glove_dict = glove_dict
self.glove_arr = glove_arr
def __str__(self):
return "%s(size = %d)" % (self.__class__.__name__, self.size)
def vector_from_word(self, word):
if word in self.glove_dict:
return self.glove_arr[self.glove_dict[word]]
else:
return torch.zeros(self.size)
def variable_from_sentence(self, sentence):
words = tokenize_sentence(sentence.lower())
tensor = torch.zeros(len(words), 1, self.size)
for wi in range(len(words)):
word = words[wi]
tensor[wi][0] = self.vector_from_word(word)
return Variable(tensor)
input_lang = GloVeLang(100)
print("input lang = %s" % input_lang)
input_lang.variable_from_sentence('turn on the light').size()
# Now we can use these Lang objects to create tensors from sentences:
def variables_from_pair(pair):
input_variable = input_lang.variable_from_sentence(pair[0])
target_variable = output_lang.variable_from_sentence(pair[1])
return (input_variable, target_variable)
def generate_training_pairs(n_iters):
pairs = []
for i in range(n_iters):
pairs.append(variables_from_pair(random_training_pair()))
return pairs