Update phrase_retrieval.py
Browse files- phrase_retrieval.py +19 -10
phrase_retrieval.py
CHANGED
@@ -28,14 +28,21 @@ logger = datasets.logging.get_logger(__name__)
|
|
28 |
|
29 |
|
30 |
_CITATION = """\
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
"""
|
32 |
|
33 |
_DESCRIPTION = """\
|
|
|
34 |
"""
|
35 |
|
36 |
-
_HOMEPAGE = ""
|
37 |
|
38 |
-
_LICENSE = "CC-BY-4.0"
|
39 |
|
40 |
_URL = "https://auburn.edu/~tmp0038/PiC/"
|
41 |
_SPLITS = {
|
@@ -65,12 +72,12 @@ class PhraseRetrieval(datasets.GeneratorBasedBuilder):
|
|
65 |
BUILDER_CONFIGS = [
|
66 |
PRConfig(
|
67 |
name=_PR_PASS,
|
68 |
-
version=datasets.Version("1.0.
|
69 |
description="The PiC Dataset for Phrase Retrieval at short passage level (~11 sentences)"
|
70 |
),
|
71 |
PRConfig(
|
72 |
name=_PR_PAGE,
|
73 |
-
version=datasets.Version("1.0.
|
74 |
description="The PiC Dataset for Phrase Retrieval at Wiki page level"
|
75 |
),
|
76 |
]
|
@@ -83,13 +90,13 @@ class PhraseRetrieval(datasets.GeneratorBasedBuilder):
|
|
83 |
"id": datasets.Value("string"),
|
84 |
"title": datasets.Value("string"),
|
85 |
"context": datasets.Value("string"),
|
86 |
-
"
|
87 |
"answers": datasets.Sequence(
|
88 |
{
|
89 |
"text": datasets.Value("string"),
|
90 |
"answer_start": datasets.Value("int32"),
|
91 |
}
|
92 |
-
)
|
93 |
}
|
94 |
),
|
95 |
# No default supervised_keys (as we have to pass both question and context as input).
|
@@ -105,7 +112,6 @@ class PhraseRetrieval(datasets.GeneratorBasedBuilder):
|
|
105 |
)
|
106 |
|
107 |
def _split_generators(self, dl_manager):
|
108 |
-
|
109 |
urls_to_download = {
|
110 |
"train": os.path.join(_URL, self.config.name, _SPLITS["train"]),
|
111 |
"dev": os.path.join(_URL, self.config.name, _SPLITS["dev"]),
|
@@ -126,20 +132,23 @@ class PhraseRetrieval(datasets.GeneratorBasedBuilder):
|
|
126 |
with open(filepath, encoding="utf-8") as f:
|
127 |
pic_pr = json.load(f)
|
128 |
for example in pic_pr["data"]:
|
|
|
|
|
129 |
answer_starts = [answer["answer_start"] for answer in example["answers"]]
|
130 |
answers = [answer["text"] for answer in example["answers"]]
|
131 |
|
132 |
# Features currently used are "context", "question", and "answers".
|
133 |
# Others are extracted here for the ease of future expansions.
|
134 |
yield key, {
|
135 |
-
"title":
|
136 |
"context": example["context"],
|
137 |
-
"
|
138 |
"id": example["id"],
|
139 |
"answers": {
|
140 |
"answer_start": answer_starts,
|
141 |
"text": answers,
|
142 |
-
}
|
143 |
}
|
144 |
key += 1
|
145 |
|
|
|
|
28 |
|
29 |
|
30 |
_CITATION = """\
|
31 |
+
@article{pham2022PiC,
|
32 |
+
title={PiC: A Phrase-in-Context Dataset for Phrase Understanding and Semantic Search},
|
33 |
+
author={Pham, Thang M and Yoon, Seunghyun and Bui, Trung and Nguyen, Anh},
|
34 |
+
journal={arXiv preprint arXiv:2207.09068},
|
35 |
+
year={2022}
|
36 |
+
}
|
37 |
"""
|
38 |
|
39 |
_DESCRIPTION = """\
|
40 |
+
Phrase in Context is a curated benchmark for phrase understanding and semantic search, consisting of three tasks of increasing difficulty: Phrase Similarity (PS), Phrase Retrieval (PR) and Phrase Sense Disambiguation (PSD). The datasets are annotated by 13 linguistic experts on Upwork and verified by two groups: ~1000 AMT crowdworkers and another set of 5 linguistic experts. PiC benchmark is distributed under CC-BY-NC 4.0.
|
41 |
"""
|
42 |
|
43 |
+
_HOMEPAGE = "https://phrase-in-context.github.io/"
|
44 |
|
45 |
+
_LICENSE = "CC-BY-NC-4.0"
|
46 |
|
47 |
_URL = "https://auburn.edu/~tmp0038/PiC/"
|
48 |
_SPLITS = {
|
|
|
72 |
BUILDER_CONFIGS = [
|
73 |
PRConfig(
|
74 |
name=_PR_PASS,
|
75 |
+
version=datasets.Version("1.0.2"),
|
76 |
description="The PiC Dataset for Phrase Retrieval at short passage level (~11 sentences)"
|
77 |
),
|
78 |
PRConfig(
|
79 |
name=_PR_PAGE,
|
80 |
+
version=datasets.Version("1.0.2"),
|
81 |
description="The PiC Dataset for Phrase Retrieval at Wiki page level"
|
82 |
),
|
83 |
]
|
|
|
90 |
"id": datasets.Value("string"),
|
91 |
"title": datasets.Value("string"),
|
92 |
"context": datasets.Value("string"),
|
93 |
+
"question": datasets.Value("string"),
|
94 |
"answers": datasets.Sequence(
|
95 |
{
|
96 |
"text": datasets.Value("string"),
|
97 |
"answer_start": datasets.Value("int32"),
|
98 |
}
|
99 |
+
)
|
100 |
}
|
101 |
),
|
102 |
# No default supervised_keys (as we have to pass both question and context as input).
|
|
|
112 |
)
|
113 |
|
114 |
def _split_generators(self, dl_manager):
|
|
|
115 |
urls_to_download = {
|
116 |
"train": os.path.join(_URL, self.config.name, _SPLITS["train"]),
|
117 |
"dev": os.path.join(_URL, self.config.name, _SPLITS["dev"]),
|
|
|
132 |
with open(filepath, encoding="utf-8") as f:
|
133 |
pic_pr = json.load(f)
|
134 |
for example in pic_pr["data"]:
|
135 |
+
title = example.get("title", "")
|
136 |
+
|
137 |
answer_starts = [answer["answer_start"] for answer in example["answers"]]
|
138 |
answers = [answer["text"] for answer in example["answers"]]
|
139 |
|
140 |
# Features currently used are "context", "question", and "answers".
|
141 |
# Others are extracted here for the ease of future expansions.
|
142 |
yield key, {
|
143 |
+
"title": title,
|
144 |
"context": example["context"],
|
145 |
+
"question": example["question"],
|
146 |
"id": example["id"],
|
147 |
"answers": {
|
148 |
"answer_start": answer_starts,
|
149 |
"text": answers,
|
150 |
+
}
|
151 |
}
|
152 |
key += 1
|
153 |
|
154 |
+
|