Datasets:
Inflate dataset
Browse files- .gitattributes +4 -0
- README.md +9 -0
- challenge.json +3 -0
- dev.json +3 -0
- test.json +3 -0
- train.json +3 -0
- webnlgqa.py +199 -0
.gitattributes
CHANGED
@@ -53,3 +53,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
53 |
*.jpg filter=lfs diff=lfs merge=lfs -text
|
54 |
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
55 |
*.webp filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
53 |
*.jpg filter=lfs diff=lfs merge=lfs -text
|
54 |
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
55 |
*.webp filter=lfs diff=lfs merge=lfs -text
|
56 |
+
train.json filter=lfs diff=lfs merge=lfs -text
|
57 |
+
dev.json filter=lfs diff=lfs merge=lfs -text
|
58 |
+
test.json filter=lfs diff=lfs merge=lfs -text
|
59 |
+
challenge.json filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
@@ -1,3 +1,12 @@
|
|
1 |
---
|
2 |
license: cc-by-sa-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
1 |
---
|
2 |
license: cc-by-sa-4.0
|
3 |
+
task_categories:
|
4 |
+
- conversational
|
5 |
+
- question-answering
|
6 |
+
- text-generation
|
7 |
+
tags:
|
8 |
+
- qa
|
9 |
+
- knowledge-graph
|
10 |
+
language:
|
11 |
+
- en
|
12 |
---
|
challenge.json
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dd2f4fc1e61d2e9ef9a059bb15a4cc6ef1c1c68534ee80474be53a6cf6026f4b
|
3 |
+
size 968686
|
dev.json
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0518c960195c31ad995fb0b700cb0b84ed4f6717365a1416c7c1ee143928300f
|
3 |
+
size 10345623
|
test.json
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1390503972269520e74edc2f3489114c488b05ab3310be36cc2e3f0ead92ccd4
|
3 |
+
size 11856922
|
train.json
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5f77a46ef327cf8f0a56cefd1bbd170bc64551c7fbba10db9e289e26f1f80acd
|
3 |
+
size 82137230
|
webnlgqa.py
ADDED
@@ -0,0 +1,199 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import zipfile
|
3 |
+
import json
|
4 |
+
import base64
|
5 |
+
import sys
|
6 |
+
import traceback
|
7 |
+
|
8 |
+
import datasets
|
9 |
+
|
10 |
+
_CITATION = """\
|
11 |
+
@inproceedings{lecorve2022sparql2text,
|
12 |
+
title={SPARQL-to-Text Question Generation for Knowledge-Based Conversational Applications},
|
13 |
+
author={Lecorv\'e, Gw\'enol\'e and Veyret, Morgan and Brabant, Quentin and Rojas-Barahona, Lina M.},
|
14 |
+
journal={Proceedings of the Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the International Joint Conference on Natural Language Processing (AACL-IJCNLP)},
|
15 |
+
year={2022}
|
16 |
+
}
|
17 |
+
"""
|
18 |
+
|
19 |
+
_HOMEPAGE = ""
|
20 |
+
|
21 |
+
_URLS = {
|
22 |
+
"train": "train.json",
|
23 |
+
"dev": "dev.json",
|
24 |
+
"test": "test.json",
|
25 |
+
"challenge": "challenge.json"
|
26 |
+
}
|
27 |
+
|
28 |
+
_DESCRIPTION = """\
|
29 |
+
Augmented version of WebNLG v3.0 English with follow-up SPARQL queries with their associated answer(s). A small portion of it also contains natural language questions associated with the queries.
|
30 |
+
"""
|
31 |
+
|
32 |
+
class WebNLGQA(datasets.GeneratorBasedBuilder):
|
33 |
+
"""
|
34 |
+
WebNLG-QA: Augmented version of WebNLG v3.0 English with follow-up SPARQL queries with their associated answer(s). A small portion of it also contains natural language questions associated with the queries.
|
35 |
+
"""
|
36 |
+
|
37 |
+
VERSION = datasets.Version("1.0.0")
|
38 |
+
|
39 |
+
def _info(self):
|
40 |
+
return datasets.DatasetInfo(
|
41 |
+
# This is the description that will appear on the datasets page.
|
42 |
+
description=_DESCRIPTION,
|
43 |
+
# datasets.features.FeatureConnectors
|
44 |
+
features=datasets.Features(
|
45 |
+
{
|
46 |
+
"category": datasets.Value("string"),
|
47 |
+
"size": datasets.Value("int32"),
|
48 |
+
"id": datasets.Value("string"),
|
49 |
+
"eid": datasets.Value("string"),
|
50 |
+
"original_triple_sets": [
|
51 |
+
{"subject": datasets.Value("string"),
|
52 |
+
"property": datasets.Value("string"),
|
53 |
+
"object": datasets.Value("string")}
|
54 |
+
],
|
55 |
+
"modified_triple_sets": [
|
56 |
+
{"subject": datasets.Value("string"),
|
57 |
+
"property": datasets.Value("string"),
|
58 |
+
"object": datasets.Value("string")}
|
59 |
+
],
|
60 |
+
"shape": datasets.Value("string"),
|
61 |
+
"shape_type": datasets.Value("string"),
|
62 |
+
"lex": datasets.Sequence(
|
63 |
+
{
|
64 |
+
"comment": datasets.Value("string"),
|
65 |
+
"lid": datasets.Value("string"),
|
66 |
+
"text": datasets.Value("string"),
|
67 |
+
"lang": datasets.Value("string"),
|
68 |
+
}
|
69 |
+
),
|
70 |
+
"test_category": datasets.Value("string"),
|
71 |
+
"dbpedia_links": datasets.Sequence(datasets.Value("string")),
|
72 |
+
"links": datasets.Sequence(datasets.Value("string")),
|
73 |
+
"graph": [
|
74 |
+
[datasets.Value("string")]
|
75 |
+
],
|
76 |
+
"main_entity": datasets.Value("string"),
|
77 |
+
"mappings": [
|
78 |
+
{
|
79 |
+
"modified": datasets.Value("string"),
|
80 |
+
"readable": datasets.Value("string"),
|
81 |
+
"graph": datasets.Value("string")
|
82 |
+
}
|
83 |
+
],
|
84 |
+
"dialogue": [
|
85 |
+
{
|
86 |
+
"question": [ {
|
87 |
+
"source": datasets.Value("string"),
|
88 |
+
"text": datasets.Value("string")
|
89 |
+
}],
|
90 |
+
"graph_query": datasets.Value("string"),
|
91 |
+
"readable_query": datasets.Value("string"),
|
92 |
+
"graph_answer": [
|
93 |
+
datasets.Value("string")
|
94 |
+
],
|
95 |
+
"readable_answer": [
|
96 |
+
datasets.Value("string")
|
97 |
+
],
|
98 |
+
"type": [ datasets.Value("string") ]
|
99 |
+
}
|
100 |
+
]
|
101 |
+
}
|
102 |
+
),
|
103 |
+
# If there's a common (input, target) tuple from the features,
|
104 |
+
# specify them here. They'll be used if as_supervised=True in
|
105 |
+
# builder.as_dataset
|
106 |
+
supervised_keys=None,
|
107 |
+
# Homepage of the dataset for documentation
|
108 |
+
homepage=_HOMEPAGE,
|
109 |
+
citation=_CITATION,
|
110 |
+
)
|
111 |
+
|
112 |
+
def _split_generators(self, dl_manager):
|
113 |
+
"""Returns SplitGenerators."""
|
114 |
+
# Downloads the data and defines the splits
|
115 |
+
# dl_manager is a datasets.download.DownloadManager that can be used to
|
116 |
+
# download and extract URLs
|
117 |
+
paths = dl_manager.download_and_extract(_URLS)
|
118 |
+
return [
|
119 |
+
datasets.SplitGenerator(
|
120 |
+
name=datasets.Split.TRAIN,
|
121 |
+
gen_kwargs={"filepath": paths['train'],
|
122 |
+
"split": "train"}
|
123 |
+
),
|
124 |
+
datasets.SplitGenerator(
|
125 |
+
name=datasets.Split.VALIDATION,
|
126 |
+
gen_kwargs={"filepath": paths['dev'],
|
127 |
+
"split": "dev"}
|
128 |
+
),
|
129 |
+
datasets.SplitGenerator(
|
130 |
+
name=datasets.Split.TEST,
|
131 |
+
gen_kwargs={"filepath": paths['test'],
|
132 |
+
"split": "test"}
|
133 |
+
),
|
134 |
+
datasets.SplitGenerator(
|
135 |
+
name="challenge",
|
136 |
+
gen_kwargs={"filepath": paths['challenge'],
|
137 |
+
"split": "challenge"}
|
138 |
+
)
|
139 |
+
]
|
140 |
+
|
141 |
+
|
142 |
+
def _generate_examples(self, filepath, split):
|
143 |
+
"""Yields examples."""
|
144 |
+
|
145 |
+
def transform_sample(original_sample):
|
146 |
+
transformed_sample = {
|
147 |
+
"category": "",
|
148 |
+
"size": -1,
|
149 |
+
"id": "",
|
150 |
+
"eid": "",
|
151 |
+
"original_triple_sets": [],
|
152 |
+
"modified_triple_sets": [],
|
153 |
+
"shape": "",
|
154 |
+
"shape_type": "",
|
155 |
+
"lex": [],
|
156 |
+
"test_category": "",
|
157 |
+
"dbpedia_links": [],
|
158 |
+
"links": [],
|
159 |
+
"graph": [],
|
160 |
+
"main_entity": "",
|
161 |
+
"mappings": [],
|
162 |
+
"dialogue": []
|
163 |
+
}
|
164 |
+
|
165 |
+
for (old_key, new_key) in [("modifiedtripleset", "modified_triple_sets"), ("originaltriplesets", "original_triple_sets"), ("dbpedialinks", "dbpedia_links"), ("lexicalisations", "lex"), ("xml_id", "eid")]:
|
166 |
+
original_sample[new_key] = original_sample[old_key]
|
167 |
+
del original_sample[old_key]
|
168 |
+
|
169 |
+
original_sample["original_triple_sets"] = original_sample["original_triple_sets"]["originaltripleset"][0]
|
170 |
+
|
171 |
+
for l in original_sample["lex"]:
|
172 |
+
l["lid"] = l["xml_id"]
|
173 |
+
del l["xml_id"]
|
174 |
+
l["text"] = l["lex"]
|
175 |
+
del l["lex"]
|
176 |
+
|
177 |
+
for turn in original_sample["dialogue"]:
|
178 |
+
if "question" in turn:
|
179 |
+
old_format = turn["question"]
|
180 |
+
new_format = []
|
181 |
+
for source, text in old_format.items():
|
182 |
+
new_format.append({"source": source, "text": text})
|
183 |
+
turn["question"] = new_format
|
184 |
+
|
185 |
+
|
186 |
+
for k in transformed_sample:
|
187 |
+
if k in original_sample:
|
188 |
+
transformed_sample[k] = original_sample[k]
|
189 |
+
# transformed_sample.update(original_sample)
|
190 |
+
|
191 |
+
return transformed_sample
|
192 |
+
|
193 |
+
# Yields (key, example) tuples from the dataset
|
194 |
+
with open(filepath,'r') as f:
|
195 |
+
data = json.load(f)
|
196 |
+
key = 0
|
197 |
+
for it in data:
|
198 |
+
yield key, transform_sample(it)
|
199 |
+
key += 1
|