-
Notifications
You must be signed in to change notification settings - Fork 0
/
neural_input_preprocessing_abstracts.py
191 lines (156 loc) · 6.92 KB
/
neural_input_preprocessing_abstracts.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
import json
import argparse
import os
from transformers import AutoTokenizer
from tqdm import tqdm
parser = argparse.ArgumentParser()
parser.add_argument("--DATA_DIR", type=str, default="data/sysrev-seed-collection")
parser.add_argument("--prompt_type", type=str, default='openai', help="openai or alpaca or biogpt or real_seed")
parser.add_argument("--tokeniser", type=str, default="dmis-lab/biobert-v1.1")
parser.add_argument("--collection_file", type=str, default="data/sysrev-seed-collection/all.jsonl")
parser.add_argument("--cache_dir", type=str, default="./cache")
args = parser.parse_args()
input_file = os.path.join(args.DATA_DIR, "preprocessed_input.jsonl")
collection_file = args.collection_file
collection_out_folder = os.path.join(args.collection_file.replace('/all.jsonl', ""), "neural_collections")
collection_index = os.path.join(collection_out_folder, args.tokeniser.replace("/", "_") + ".jsonl")
if not os.path.exists(collection_out_folder):
os.makedirs(collection_out_folder)
passage_title_dict = {}
passage_abstract_dict = {}
passage_tiab_dict = {}
tokenizer = AutoTokenizer.from_pretrained(args.tokeniser, use_fast=True, cache_dir=args.cache_dir)
# below is for creating collection index
if not os.path.exists(collection_index):
with open(collection_index, 'w') as fw:
with open(collection_file) as f:
for line in tqdm(f):
current_dict = json.loads(line)
pid = current_dict["pmid"]
title = current_dict["title"]
abstract = current_dict["abstract"]
title_tokenized = tokenizer.encode(
title,
add_special_tokens=False,
max_length=64,
truncation=True,
padding="max_length"
)
t_ab_tokenized = tokenizer.encode(
title + "[SEP]" + abstract,
add_special_tokens=False,
max_length=448,
truncation=True,
padding="max_length")
ab_tokenized = tokenizer.encode(
abstract,
add_special_tokens=False,
)
current_dict["title_tokenized"] = title_tokenized
current_dict["title_abstract_tokenized"] = t_ab_tokenized
current_dict["abstract_tokenized"] = ab_tokenized
fw.write(json.dumps(current_dict) + "\n")
passage_title_dict[pid] = title_tokenized
passage_abstract_dict[pid] = ab_tokenized
passage_tiab_dict[pid] = t_ab_tokenized
else:
with open(collection_index) as f:
for line in tqdm(f):
current_dict = json.loads(line)
pid = current_dict["pmid"]
title = current_dict["title"] # this is tokenized
if "abstract" in current_dict:
abstract = current_dict["abstract"] # this is not tokenized
passage_abstract_dict[pid] = abstract
tiab = current_dict["title_abstract"]
passage_title_dict[pid] = title
passage_tiab_dict[pid] = tiab
prompt_dict = {}
if args.prompt_type == "real_seed":
seed_file = os.path.join(args.DATA_DIR, f'preprocessed_input.jsonl')
with open(seed_file) as f:
for line in tqdm(f):
current_dict = json.loads(line)
qid = current_dict["id"]
title = current_dict["title"]
seed_studies = current_dict["seed_studies"]
prompt_dict[qid] = {}
if qid not in prompt_dict:
prompt_dict[qid] = {}
for index, s_id in enumerate(seed_studies):
if s_id not in passage_tiab_dict:
continue
abstract = passage_abstract_dict[s_id]
query = title + " " + abstract
prompt_dict[qid][index] = query
else:
input_type = args.prompt_type.split("_")[0]
prompt_seed_file = os.path.join(args.DATA_DIR, f'generated_from_{args.prompt_type}.jsonl')
with open(prompt_seed_file) as f:
for line in tqdm(f):
current_dict = json.loads(line)
qid = current_dict["id"]
if "title" in current_dict:
original_info = current_dict["title"]
elif "abstract" in current_dict:
original_info = current_dict["abstract"]
else:
original_info = current_dict["boolean"]
generated_query = current_dict["generated_query"]
prompt_dict[qid] = generated_query
# below is for creating query index and read inputfile
query_dict = {}
passage_mapping = {}
with open(input_file) as f:
for line in tqdm(f):
current_dict = json.loads(line)
qid = current_dict["id"]
input_type = args.prompt_type.split("_")[0]
query_list = prompt_dict[qid]
query_dict[qid] = {}
if "title" in input_type:
for index, query in enumerate(query_list):
query_dict[qid][index] = tokenizer.encode(
query,
add_special_tokens=False,
max_length=64,
truncation=True,
padding="max_length")
search_docs = current_dict["search_docs"]
passage_mapping[qid] = search_docs
else:
for index, query in enumerate(query_list):
query_dict[qid][index] = tokenizer.encode(
query,
add_special_tokens=False,
max_length=256,
truncation=True,
padding="max_length")
search_docs = current_dict["search_docs"]
passage_mapping[qid] = search_docs
# bwlow is for output
if args.prompt_type.split("_")[0].endswith("s"):
redefined_prompt = args.prompt_type.split("_", 1)[0][:-1] + "_" + args.prompt_type.split("_", 1)[1]
out_folder = os.path.join(args.DATA_DIR, "neural_inputs", redefined_prompt + "_multi", args.tokeniser)
if not os.path.exists(out_folder):
os.makedirs(out_folder)
out_run_file = os.path.join(out_folder, 'run.jsonl')
out_run_tsv_file = os.path.join(out_folder, 'run.tsv')
out_runs = []
out_runs_tsv = []
for qid, q_dict in tqdm(query_dict.items()):
for index, query in q_dict.items():
current_dict = {}
current_dict["qid"] = qid + "_" + str(index)
current_dict["qry"] = query
if qid in passage_mapping:
for pid in passage_mapping[qid]:
if pid not in passage_tiab_dict:
continue
current_dict["pid"] = pid
current_dict["psg"] = passage_tiab_dict[pid]
out_runs.append(json.dumps(current_dict) + "\n")
out_runs_tsv.append(f"{qid}_{index}\t{pid}\n")
with open(out_run_file, "w") as out_rerank, open(out_run_tsv_file, "w") as out_rerank_tsv:
out_rerank_tsv.writelines(out_runs_tsv)
out_rerank.writelines(out_runs)