-
Notifications
You must be signed in to change notification settings - Fork 9
/
highlight_sentence.py
143 lines (114 loc) · 4.66 KB
/
highlight_sentence.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
from typing import List, Dict, Callable, Tuple, Any
import argparse
import time
import random
from dataclasses import dataclass, field
from pathlib import Path
import json
from tqdm.auto import tqdm
from allennlp.predictors.predictor import Predictor
import spacy
random.seed(0)
nlp = spacy.load("en")
def load_json(infile: Path):
with open(infile) as fid:
return json.load(fid)
def save_json(dataset: Any, outfile: Path):
with open(outfile, "w") as fid:
return json.dump(dataset, fid, indent=2)
Highlighter = Callable[[str], List[str]]
@dataclass
class POS:
name: str
span: Tuple[int, int]
@dataclass
class POSHighlighter:
predictor: Predictor = field(
default_factory=lambda: Predictor.from_path(
"/gpfsdswork/projects/rech/vuo/uok79zh/.allennlp/elmo/",
cuda_device=0
# "https://storage.googleapis.com/allennlp-public-models/elmo-constituency-parser-2020.02.10.tar.gz"
)
)
# turn is causing a lot of confusion to the parser
forbidden: Tuple[str, ...] = ("next", "then", "there", "just")
pos: Tuple[str, ...] = ("NN", "NNS", "NNP", "NNPS", "RB", "RBR", "RBS")
def __post_init__(self):
self._tokenizer = self.predictor._tokenizer
def _retrieve_pos(self, sentence: str, tree: Dict, pos: int = 0) -> List[POS]:
"""
Return a dictionary with noun phrases and the spanning positions
"""
inner_pos = 0
part_of_speeches: List[POS] = []
for children in tree["children"]:
next_char = len(children["word"]) + pos + inner_pos
if next_char < len(sentence) and sentence[next_char].isspace():
children["word"] += " "
proposal = children["word"].strip()
if children["nodeType"] in self.pos and proposal not in self.forbidden:
start = tree["word"][inner_pos:].find(proposal) + pos + inner_pos
end = start + len(proposal) - 1
part_of_speeches.append(POS(proposal, (start, end)))
inner_pos += len(children["word"])
continue
if "children" in children:
start = tree["word"][inner_pos:].find(children["word"]) + pos + inner_pos
part_of_speeches += self._retrieve_pos(sentence, children, pos=start)
inner_pos += len(children["word"])
return part_of_speeches
def __call__(self, sentence: str) -> List[str]:
tokens = self._tokenizer.tokenize(sentence.lower().rstrip())
fake_sentence = " ".join([str(token) for token in tokens])
preds = self.predictor.predict(sentence=fake_sentence) # type: ignore
pos = self._retrieve_pos(sentence, preds["hierplane_tree"]["root"])
# sort the pos by start span
pos = sorted(pos, key=lambda item: item.span[0])
return [p.name for p in pos]
def highlight_dataset(
dataset: List[Dict],
highlighter: Highlighter,
num_samples: int,
) -> None:
counter = 0
for item in tqdm(dataset):
if "highlights" not in item:
item["highlights"] = [highlighter(instr) for instr in item["instructions"]]
if "perturbations" in item and "perturbation_highlights" not in item:
item["perturbation_highlights"] = []
for instructions in item["perturbations"]:
item["perturbation_highlights"].append(
[highlighter(instr) for instr in instructions]
)
counter += 1
if num_samples > -1 and counter >= num_samples:
return
def get_parser():
parser = argparse.ArgumentParser()
parser.add_argument("--infile", type=Path)
parser.add_argument("--outfile", type=Path)
parser.add_argument("--num-samples", type=int, default=-1)
parser.add_argument("--id", type=int, default=0)
parser.add_argument("--num-procs", type=int, default=1)
parser.add_argument(
"--pos",
type=str,
nargs="+",
default=("NN", "NNS", "NNP", "NNPS", "RB", "RBR", "RBS"),
)
return parser.parse_args()
if __name__ == "__main__":
args = get_parser()
dataset = load_json(args.infile)
highlighter = POSHighlighter(pos=args.pos)
highlight_dataset(dataset[args.id :: args.num_procs], highlighter, args.num_samples) # type: ignore
print("Exporting to", args.outfile)
lock = args.outfile.parent / f"{args.outfile.stem}.lock"
while lock.is_file():
print(f"lock found at {lock}. sleep 1 sec")
time.sleep(1)
lock.touch()
uptodate = load_json(args.infile)
uptodate[args.id :: args.num_procs] = dataset[args.id :: args.num_procs]
save_json(uptodate, args.outfile)
lock.unlink()