Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
10K - 100K
License:
Delete _logicnlg.py
Browse files- _logicnlg.py +0 -84
_logicnlg.py
DELETED
@@ -1,84 +0,0 @@
|
|
1 |
-
#!/usr/bin/env python3
|
2 |
-
|
3 |
-
"""
|
4 |
-
The script used to load the dataset from the original source.
|
5 |
-
"""
|
6 |
-
|
7 |
-
import json
|
8 |
-
import datasets
|
9 |
-
import glob
|
10 |
-
import os
|
11 |
-
|
12 |
-
_CITATION = """\
|
13 |
-
@inproceedings{chen2020logical,
|
14 |
-
title={Logical Natural Language Generation from Open-Domain Tables},
|
15 |
-
author={Chen, Wenhu and Chen, Jianshu and Su, Yu and Chen, Zhiyu and Wang, William Yang},
|
16 |
-
booktitle={Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
|
17 |
-
pages={7929--7942},
|
18 |
-
year={2020}
|
19 |
-
}
|
20 |
-
"""
|
21 |
-
_DESCRIPTION = """\
|
22 |
-
LogicNLG is a dataset for natural language generation from open-domain tables.
|
23 |
-
LogicNLG is based on TabFact (Chen et al., 2019), which is a table-based fact-checking dataset with rich logical inferences in the annotated statements.
|
24 |
-
"""
|
25 |
-
|
26 |
-
_URL = "https://github.com/wenhuchen/LogicNLG"
|
27 |
-
_LICENSE = "MIT"
|
28 |
-
|
29 |
-
class LogicNLG(datasets.GeneratorBasedBuilder):
|
30 |
-
VERSION = "1.0.0"
|
31 |
-
def _info(self):
|
32 |
-
return datasets.DatasetInfo(
|
33 |
-
description=_DESCRIPTION,
|
34 |
-
features=datasets.Features({
|
35 |
-
'table': datasets.Value(dtype='large_string'),
|
36 |
-
"ref": datasets.Value(dtype='string'),
|
37 |
-
"linked_columns": datasets.Value(dtype='string'),
|
38 |
-
"title": datasets.Value(dtype='string'),
|
39 |
-
"template": datasets.Value(dtype='string'),
|
40 |
-
"table_id": datasets.Value(dtype='string'),
|
41 |
-
}),
|
42 |
-
supervised_keys=None,
|
43 |
-
homepage="https://wenhuchen.github.io/logicnlg.github.io/",
|
44 |
-
citation=_CITATION,
|
45 |
-
license=_LICENSE,
|
46 |
-
)
|
47 |
-
|
48 |
-
def _split_generators(self, dl_manager):
|
49 |
-
"""Returns SplitGenerators."""
|
50 |
-
return [
|
51 |
-
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": "data", "split" : "train"}),
|
52 |
-
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": "data", "split" : "dev"}),
|
53 |
-
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": "data", "split" : "test"}),
|
54 |
-
]
|
55 |
-
|
56 |
-
def _generate_examples(self, filepath, split):
|
57 |
-
filename = split if split != "dev" else "val"
|
58 |
-
data = []
|
59 |
-
|
60 |
-
with open(os.path.join(filepath, f"{filename}_lm.json")) as f:
|
61 |
-
j = json.load(f)
|
62 |
-
|
63 |
-
for i, (table_id, examples) in enumerate(j.items()):
|
64 |
-
table = []
|
65 |
-
with open(os.path.join(filepath, "all_csv", table_id)) as f:
|
66 |
-
for line in f.readlines():
|
67 |
-
table.append(line.rstrip("\n").split("#"))
|
68 |
-
|
69 |
-
for example in examples:
|
70 |
-
data.append({
|
71 |
-
"table": table,
|
72 |
-
"ref": example[0],
|
73 |
-
"linked_columns": example[1],
|
74 |
-
"title": example[2],
|
75 |
-
"template": example[3],
|
76 |
-
"table_id": table_id,
|
77 |
-
})
|
78 |
-
for example_idx, entry in enumerate(data):
|
79 |
-
yield example_idx, {key: str(value) for key, value in entry.items()}
|
80 |
-
|
81 |
-
|
82 |
-
if __name__ == '__main__':
|
83 |
-
dataset = datasets.load_dataset(__file__)
|
84 |
-
dataset.push_to_hub("kasnerz/logicnlg")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|