kasnerz commited on
Commit
2649762
·
1 Parent(s): 8d436d3

Upload _numericnlg.py

Browse files
Files changed (1) hide show
  1. _numericnlg.py +88 -0
_numericnlg.py ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python3
2
+
3
+ """
4
+ The script used to load the dataset from the original source.
5
+ """
6
+
7
+
8
+ import json
9
+ import datasets
10
+ import glob
11
+ import os
12
+
13
+ _CITATION = """\
14
+ @inproceedings{suadaa-etal-2021-towards,
15
+ title = "Towards Table-to-Text Generation with Numerical Reasoning",
16
+ author = "Suadaa, Lya Hulliyyatus and
17
+ Kamigaito, Hidetaka and
18
+ Funakoshi, Kotaro and
19
+ Okumura, Manabu and
20
+ Takamura, Hiroya",
21
+ booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
22
+ month = aug,
23
+ year = "2021",
24
+ address = "Online",
25
+ publisher = "Association for Computational Linguistics",
26
+ url = "https://aclanthology.org/2021.acl-long.115",
27
+ doi = "10.18653/v1/2021.acl-long.115",
28
+ pages = "1451--1465"
29
+ }
30
+ """
31
+ _DESCRIPTION = """\
32
+ NumericNLG is a dataset for table-totext generation focusing on numerical reasoning.
33
+ The dataset consists of textual descriptions of numerical tables from scientific papers.
34
+ """
35
+
36
+ _URL = "https://github.com/titech-nlp/numeric-nlg"
37
+ _LICENSE = "CC BY-SA 4.0"
38
+
39
+ class NumericNLG(datasets.GeneratorBasedBuilder):
40
+ VERSION = "1.0.0"
41
+ def _info(self):
42
+ return datasets.DatasetInfo(
43
+ description=_DESCRIPTION,
44
+ features=datasets.Features({
45
+ "table_id_paper": datasets.Value(dtype='string'),
46
+ "caption": datasets.Value(dtype='string'),
47
+ "row_header_level" : datasets.Value(dtype='int32'),
48
+ "row_headers" : datasets.Value(dtype='large_string'),
49
+ "column_header_level": datasets.Value(dtype='int32'),
50
+ "column_headers" : datasets.Value(dtype='large_string'),
51
+ "contents" : datasets.Value(dtype='large_string'),
52
+ "metrics_loc" : datasets.Value(dtype='string'),
53
+ "metrics_type" : datasets.Value(dtype='large_string'),
54
+ "target_entity": datasets.Value(dtype='large_string'),
55
+ "table_html_clean": datasets.Value(dtype='large_string'),
56
+ "table_name": datasets.Value(dtype='string'),
57
+ "table_id": datasets.Value(dtype='string'),
58
+ "paper_id": datasets.Value(dtype='string'),
59
+ "page_no": datasets.Value(dtype='int32'),
60
+ "dir": datasets.Value(dtype='string'),
61
+ "valid": datasets.Value(dtype='int32'),
62
+ }),
63
+ supervised_keys=None,
64
+ homepage="https://github.com/titech-nlp/numeric-nlg",
65
+ citation=_CITATION,
66
+ license=_LICENSE,
67
+ )
68
+
69
+ def _split_generators(self, dl_manager):
70
+ """Returns SplitGenerators."""
71
+ return [
72
+ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": "data", "split" : "train"}),
73
+ datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": "data", "split" : "dev"}),
74
+ datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": "data", "split" : "test"}),
75
+ ]
76
+
77
+ def _generate_examples(self, filepath, split):
78
+ filename = split if split != "dev" else "val"
79
+
80
+ with open(os.path.join(filepath, f"table_{filename}.json")) as f:
81
+ j = json.load(f)
82
+
83
+ for example_idx, entry in enumerate(j):
84
+ yield example_idx, {key: str(value) for key, value in entry.items()}
85
+
86
+ if __name__ == '__main__':
87
+ dataset = datasets.load_dataset(__file__)
88
+ dataset.push_to_hub("kasnerz/numericnlg")