Commit
•
6dc77db
1
Parent(s):
c4f4109
Delete loading script
Browse files- multi_nli.py +0 -118
multi_nli.py
DELETED
@@ -1,118 +0,0 @@
|
|
1 |
-
# coding=utf-8
|
2 |
-
# Copyright 2020 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
|
3 |
-
#
|
4 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
-
# you may not use this file except in compliance with the License.
|
6 |
-
# You may obtain a copy of the License at
|
7 |
-
#
|
8 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
-
#
|
10 |
-
# Unless required by applicable law or agreed to in writing, software
|
11 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
-
# See the License for the specific language governing permissions and
|
14 |
-
# limitations under the License.
|
15 |
-
|
16 |
-
# Lint as: python3
|
17 |
-
"""The Multi-Genre NLI Corpus."""
|
18 |
-
|
19 |
-
|
20 |
-
import json
|
21 |
-
import os
|
22 |
-
|
23 |
-
import datasets
|
24 |
-
|
25 |
-
|
26 |
-
_CITATION = """\
|
27 |
-
@InProceedings{N18-1101,
|
28 |
-
author = {Williams, Adina
|
29 |
-
and Nangia, Nikita
|
30 |
-
and Bowman, Samuel},
|
31 |
-
title = {A Broad-Coverage Challenge Corpus for
|
32 |
-
Sentence Understanding through Inference},
|
33 |
-
booktitle = {Proceedings of the 2018 Conference of
|
34 |
-
the North American Chapter of the
|
35 |
-
Association for Computational Linguistics:
|
36 |
-
Human Language Technologies, Volume 1 (Long
|
37 |
-
Papers)},
|
38 |
-
year = {2018},
|
39 |
-
publisher = {Association for Computational Linguistics},
|
40 |
-
pages = {1112--1122},
|
41 |
-
location = {New Orleans, Louisiana},
|
42 |
-
url = {http://aclweb.org/anthology/N18-1101}
|
43 |
-
}
|
44 |
-
"""
|
45 |
-
|
46 |
-
_DESCRIPTION = """\
|
47 |
-
The Multi-Genre Natural Language Inference (MultiNLI) corpus is a
|
48 |
-
crowd-sourced collection of 433k sentence pairs annotated with textual
|
49 |
-
entailment information. The corpus is modeled on the SNLI corpus, but differs in
|
50 |
-
that covers a range of genres of spoken and written text, and supports a
|
51 |
-
distinctive cross-genre generalization evaluation. The corpus served as the
|
52 |
-
basis for the shared task of the RepEval 2017 Workshop at EMNLP in Copenhagen.
|
53 |
-
"""
|
54 |
-
|
55 |
-
|
56 |
-
class MultiNli(datasets.GeneratorBasedBuilder):
|
57 |
-
"""MultiNLI: The Stanford Question Answering Dataset. Version 1.1."""
|
58 |
-
|
59 |
-
def _info(self):
|
60 |
-
return datasets.DatasetInfo(
|
61 |
-
description=_DESCRIPTION,
|
62 |
-
features=datasets.Features(
|
63 |
-
{
|
64 |
-
"promptID": datasets.Value("int32"),
|
65 |
-
"pairID": datasets.Value("string"),
|
66 |
-
"premise": datasets.Value("string"),
|
67 |
-
"premise_binary_parse": datasets.Value("string"), # parses in unlabeled binary-branching format
|
68 |
-
"premise_parse": datasets.Value("string"), # sentence as parsed by the Stanford PCFG Parser 3.5.2
|
69 |
-
"hypothesis": datasets.Value("string"),
|
70 |
-
"hypothesis_binary_parse": datasets.Value("string"), # parses in unlabeled binary-branching format
|
71 |
-
"hypothesis_parse": datasets.Value(
|
72 |
-
"string"
|
73 |
-
), # sentence as parsed by the Stanford PCFG Parser 3.5.2
|
74 |
-
"genre": datasets.Value("string"),
|
75 |
-
"label": datasets.features.ClassLabel(names=["entailment", "neutral", "contradiction"]),
|
76 |
-
}
|
77 |
-
),
|
78 |
-
# No default supervised_keys (as we have to pass both premise
|
79 |
-
# and hypothesis as input).
|
80 |
-
supervised_keys=None,
|
81 |
-
homepage="https://www.nyu.edu/projects/bowman/multinli/",
|
82 |
-
citation=_CITATION,
|
83 |
-
)
|
84 |
-
|
85 |
-
def _split_generators(self, dl_manager):
|
86 |
-
|
87 |
-
downloaded_dir = dl_manager.download_and_extract("https://cims.nyu.edu/~sbowman/multinli/multinli_1.0.zip")
|
88 |
-
mnli_path = os.path.join(downloaded_dir, "multinli_1.0")
|
89 |
-
train_path = os.path.join(mnli_path, "multinli_1.0_train.jsonl")
|
90 |
-
matched_validation_path = os.path.join(mnli_path, "multinli_1.0_dev_matched.jsonl")
|
91 |
-
mismatched_validation_path = os.path.join(mnli_path, "multinli_1.0_dev_mismatched.jsonl")
|
92 |
-
|
93 |
-
return [
|
94 |
-
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": train_path}),
|
95 |
-
datasets.SplitGenerator(name="validation_matched", gen_kwargs={"filepath": matched_validation_path}),
|
96 |
-
datasets.SplitGenerator(name="validation_mismatched", gen_kwargs={"filepath": mismatched_validation_path}),
|
97 |
-
]
|
98 |
-
|
99 |
-
def _generate_examples(self, filepath):
|
100 |
-
"""Generate mnli examples"""
|
101 |
-
|
102 |
-
with open(filepath, encoding="utf-8") as f:
|
103 |
-
for id_, row in enumerate(f):
|
104 |
-
data = json.loads(row)
|
105 |
-
if data["gold_label"] == "-":
|
106 |
-
continue
|
107 |
-
yield id_, {
|
108 |
-
"promptID": data["promptID"],
|
109 |
-
"pairID": data["pairID"],
|
110 |
-
"premise": data["sentence1"],
|
111 |
-
"premise_binary_parse": data["sentence1_binary_parse"],
|
112 |
-
"premise_parse": data["sentence1_parse"],
|
113 |
-
"hypothesis": data["sentence2"],
|
114 |
-
"hypothesis_binary_parse": data["sentence2_binary_parse"],
|
115 |
-
"hypothesis_parse": data["sentence2_parse"],
|
116 |
-
"genre": data["genre"],
|
117 |
-
"label": data["gold_label"],
|
118 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|