-
Notifications
You must be signed in to change notification settings - Fork 34
/
Copy pathgenerate_paraphrases.py
187 lines (155 loc) · 7.37 KB
/
generate_paraphrases.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
import torch, time, sys, argparse, os, codecs, h5py, cPickle, csv
import numpy as np
from torch.autograd import Variable
from nltk import ParentedTree
from train_scpn import SCPN
from train_parse_generator import ParseNet
from subwordnmt.apply_bpe import BPE, read_vocabulary
from scpn_utils import deleaf, parse_indexify_transformations
reload(sys)
sys.setdefaultencoding('utf8')
# 10 frequent templates
templates = [
'( ROOT ( S ( NP ) ( VP ) ( . ) ) ) EOP',
'( ROOT ( S ( VP ) ( . ) ) ) EOP',
'( ROOT ( NP ( NP ) ( . ) ) ) EOP',
'( ROOT ( FRAG ( SBAR ) ( . ) ) ) EOP',
'( ROOT ( S ( S ) ( , ) ( CC ) ( S ) ( . ) ) ) EOP',
'( ROOT ( S ( LST ) ( VP ) ( . ) ) ) EOP',
'( ROOT ( SBARQ ( WHADVP ) ( SQ ) ( . ) ) ) EOP',
'( ROOT ( S ( PP ) ( , ) ( NP ) ( VP ) ( . ) ) ) EOP',
'( ROOT ( S ( ADVP ) ( NP ) ( VP ) ( . ) ) ) EOP',
'( ROOT ( S ( SBAR ) ( , ) ( NP ) ( VP ) ( . ) ) ) EOP'
]
def reverse_bpe(sent):
x = []
cache = ''
for w in sent:
if w.endswith('@@'):
cache += w.replace('@@', '')
elif cache != '':
x.append(cache + w)
cache = ''
else:
x.append(w)
return ' '.join(x)
# encode sentences and parses for targeted paraphrasing
def encode_data(out_file):
fn = ['idx', 'template', 'generated_parse', 'sentence']
ofile = codecs.open(out_file, 'w', 'utf-8')
out = csv.DictWriter(ofile, delimiter='\t', fieldnames=fn)
out.writerow(dict((x,x) for x in fn))
# read parsed data
infile = codecs.open(args.parsed_input_file, 'r', 'utf-8')
inrdr = csv.DictReader(infile, delimiter='\t')
# loop over sentences and transform them
for d_idx, ex in enumerate(inrdr):
stime = time.time()
ssent = ' '.join(ex['tokens'].split())
seg_sent = bpe.segment(ssent.lower()).split()
# write gold sentence
out.writerow({'idx': ex['idx'],
'template':'GOLD', 'generated_parse':ex['parse'],
'sentence':reverse_bpe(seg_sent)})
# encode sentence using pp_vocab, leave one word for EOS
seg_sent = [pp_vocab[w] for w in seg_sent if w in pp_vocab]
# add EOS
seg_sent.append(pp_vocab['EOS'])
torch_sent = Variable(torch.from_numpy(np.array(seg_sent, dtype='int32')).long().cuda())
torch_sent_len = torch.from_numpy(np.array([len(seg_sent)], dtype='int32')).long().cuda()
# encode parse using parse vocab
parse_tree = ParentedTree.fromstring(ex['parse'].strip())
parse_tree = deleaf(parse_tree)
np_parse = np.array([parse_gen_voc[w] for w in parse_tree], dtype='int32')
torch_parse = Variable(torch.from_numpy(np_parse).long().cuda())
torch_parse_len = torch.from_numpy(np.array([len(parse_tree)], dtype='int32')).long().cuda()
# generate full parses from templates
beam_dict = parse_net.batch_beam_search(torch_parse.unsqueeze(0), tp_templates,
torch_parse_len[:], tp_template_lens, parse_gen_voc['EOP'], beam_size=3, max_steps=150)
seq_lens = []
seqs = []
for b_idx in beam_dict:
prob,_,_,seq = beam_dict[b_idx][0]
seq = seq[:-1] # chop off EOP
seq_lens.append(len(seq))
seqs.append(seq)
np_parses = np.zeros((len(seqs), max(seq_lens)), dtype='int32')
for z, seq in enumerate(seqs):
np_parses[z, :seq_lens[z]] = seq
tp_parses = Variable(torch.from_numpy(np_parses).long().cuda())
tp_len = torch.from_numpy(np.array(seq_lens, dtype='int32')).long().cuda()
# generate paraphrases from parses
try:
beam_dict = net.batch_beam_search(torch_sent.unsqueeze(0), tp_parses,
torch_sent_len[:], tp_len, pp_vocab['EOS'], beam_size=3, max_steps=40)
for b_idx in beam_dict:
prob,_,_,seq = beam_dict[b_idx][0]
gen_parse = ' '.join([rev_label_voc[z] for z in seqs[b_idx]])
gen_sent = ' '.join([rev_pp_vocab[w] for w in seq[:-1]])
out.writerow({'idx': ex['idx'],
'template':templates[b_idx], 'generated_parse':gen_parse,
'sentence':reverse_bpe(gen_sent.split())})
except:
print 'beam search OOM'
print d_idx, time.time() - stime
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Syntactically Controlled Paraphrase Transformer')
## paraphrase model args
parser.add_argument('--gpu', type=str, default='0',
help='GPU id')
parser.add_argument('--out_file', type=str, default='scpn_ex.out',
help='paraphrase save path')
parser.add_argument('--parsed_input_file', type=str, default='data/scpn_ex.tsv',
help='parse load path')
parser.add_argument('--vocab', type=str, default='data/parse_vocab.pkl',
help='word vocabulary')
parser.add_argument('--parse_vocab', type=str, default='data/ptb_tagset.txt',
help='tag vocabulary')
parser.add_argument('--pp_model', type=str, default='models/scpn.pt',
help='paraphrase model to load')
parser.add_argument('--parse_model', type=str, default='models/parse_generator.pt',
help='model save path')
## BPE args
parser.add_argument('--bpe_codes', type=str, default='data/bpe.codes')
parser.add_argument('--bpe_vocab', type=str, default='data/vocab.txt')
parser.add_argument('--bpe_vocab_thresh', type=int, default=50)
args = parser.parse_args()
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
# load saved models
pp_model = torch.load(args.pp_model)
parse_model = torch.load(args.parse_model)
# load vocab
pp_vocab, rev_pp_vocab = cPickle.load(open(args.vocab, 'rb'))
tag_file = codecs.open(args.parse_vocab, 'r', 'utf-8')
parse_gen_voc = {}
for idx, line in enumerate(tag_file):
line = line.strip()
parse_gen_voc[line] = idx
rev_label_voc = dict((v,k) for (k,v) in parse_gen_voc.iteritems())
# load paraphrase network
pp_args = pp_model['config_args']
net = SCPN(pp_args.d_word, pp_args.d_hid, pp_args.d_nt, pp_args.d_trans,
len(pp_vocab), len(parse_gen_voc) - 1, pp_args.use_input_parse)
net.cuda()
net.load_state_dict(pp_model['state_dict'])
net.eval()
# load parse generator network
parse_args = parse_model['config_args']
parse_net = ParseNet(parse_args.d_nt, parse_args.d_hid, len(parse_gen_voc))
parse_net.cuda()
parse_net.load_state_dict(parse_model['state_dict'])
parse_net.eval()
# encode templates
template_lens = [len(x.split()) for x in templates]
np_templates = np.zeros((len(templates), max(template_lens)), dtype='int32')
for z, template in enumerate(templates):
np_templates[z, :template_lens[z]] = [parse_gen_voc[w] for w in templates[z].split()]
tp_templates = Variable(torch.from_numpy(np_templates).long().cuda())
tp_template_lens = torch.from_numpy(np.array(template_lens, dtype='int32')).long().cuda()
# instantiate BPE segmenter
bpe_codes = codecs.open(args.bpe_codes, encoding='utf-8')
bpe_vocab = codecs.open(args.bpe_vocab, encoding='utf-8')
bpe_vocab = read_vocabulary(bpe_vocab, args.bpe_vocab_thresh)
bpe = BPE(bpe_codes, '@@', bpe_vocab, None)
# paraphrase the sst!
encode_data(out_file=args.out_file)