lm commited on
Commit
8f95949
1 Parent(s): ef137ea

fast loading

Browse files
README.md CHANGED
@@ -15,6 +15,55 @@ language:
15
  - deu
16
  size_categories:
17
  - 10K<n<100K
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  ---
19
 
20
  #### Description
@@ -81,4 +130,4 @@ Make sure to credit the authors and cite relevant papers (see citation below) if
81
  pages = "154--169",
82
  abstract = "Hate speech detection models are typically evaluated on held-out test sets. However, this risks painting an incomplete and potentially misleading picture of model performance because of increasingly well-documented systematic gaps and biases in hate speech datasets. To enable more targeted diagnostic insights, recent research has thus introduced functional tests for hate speech detection models. However, these tests currently only exist for English-language content, which means that they cannot support the development of more effective models in other languages spoken by billions across the world. To help address this issue, we introduce Multilingual HateCheck (MHC), a suite of functional tests for multilingual hate speech detection models. MHC covers 34 functionalities across ten languages, which is more languages than any other hate speech dataset. To illustrate MHC{'}s utility, we train and test a high-performing multilingual hate speech detection model, and reveal critical model weaknesses for monolingual and cross-lingual applications.",
83
  }
84
- ```
 
15
  - deu
16
  size_categories:
17
  - 10K<n<100K
18
+ configs:
19
+ - config_name: default
20
+ data_files:
21
+ - path: test/*.jsonl.gz
22
+ split: test
23
+ - config_name: hin
24
+ data_files:
25
+ - path: multi-hatecheck/test/hin.jsonl.gz
26
+ split: test
27
+ - config_name: spa
28
+ data_files:
29
+ - path: multi-hatecheck/test/spa.jsonl.gz
30
+ split: test
31
+ - config_name: pol
32
+ data_files:
33
+ - path: multi-hatecheck/test/pol.jsonl.gz
34
+ split: test
35
+ - config_name: eng
36
+ data_files:
37
+ - path: multi-hatecheck/test/eng.jsonl.gz
38
+ split: test
39
+ - config_name: fra
40
+ data_files:
41
+ - path: multi-hatecheck/test/fra.jsonl.gz
42
+ split: test
43
+ - config_name: nld
44
+ data_files:
45
+ - path: multi-hatecheck/test/nld.jsonl.gz
46
+ split: test
47
+ - config_name: ita
48
+ data_files:
49
+ - path: multi-hatecheck/test/ita.jsonl.gz
50
+ split: test
51
+ - config_name: deu
52
+ data_files:
53
+ - path: multi-hatecheck/test/deu.jsonl.gz
54
+ split: test
55
+ - config_name: ara
56
+ data_files:
57
+ - path: multi-hatecheck/test/ara.jsonl.gz
58
+ split: test
59
+ - config_name: por
60
+ data_files:
61
+ - path: multi-hatecheck/test/por.jsonl.gz
62
+ split: test
63
+ - config_name: cmn
64
+ data_files:
65
+ - path: multi-hatecheck/test/cmn.jsonl.gz
66
+ split: test
67
  ---
68
 
69
  #### Description
 
130
  pages = "154--169",
131
  abstract = "Hate speech detection models are typically evaluated on held-out test sets. However, this risks painting an incomplete and potentially misleading picture of model performance because of increasingly well-documented systematic gaps and biases in hate speech datasets. To enable more targeted diagnostic insights, recent research has thus introduced functional tests for hate speech detection models. However, these tests currently only exist for English-language content, which means that they cannot support the development of more effective models in other languages spoken by billions across the world. To help address this issue, we introduce Multilingual HateCheck (MHC), a suite of functional tests for multilingual hate speech detection models. MHC covers 34 functionalities across ten languages, which is more languages than any other hate speech dataset. To illustrate MHC{'}s utility, we train and test a high-performing multilingual hate speech detection model, and reveal critical model weaknesses for monolingual and cross-lingual applications.",
132
  }
133
+ ```
ara/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
cmn/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
deu/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
eng/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
fra/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
hin/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
ita/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
multi-hatecheck.py DELETED
@@ -1,148 +0,0 @@
1
- import json
2
- import datasets
3
-
4
-
5
-
6
-
7
- _DESCRIPTION = """\
8
- Combines multilingual HateCheck datasets (10 languages, including English), by Paul Roettger.
9
- The original English dataset can be found under https://github.com/Paul/hatecheck.
10
- Datasets for other languages are found at:
11
- - https://github.com/Paul/hatecheck-arabic
12
- - https://github.com/Paul/hatecheck-mandarin
13
- - https://github.com/Paul/hatecheck-german
14
- - https://github.com/Paul/hatecheck-french
15
- - https://github.com/Paul/hatecheck-hindi
16
- - https://github.com/Paul/hatecheck-italian
17
- - https://github.com/Paul/hatecheck-dutch
18
- - https://github.com/Paul/hatecheck-portuguese
19
- - https://github.com/Paul/hatecheck-spanish
20
- Make sure to credit the authors and cite relevant papers if you use these datasets.
21
- """
22
- _CITATION = """\
23
- @inproceedings{rottger-etal-2021-hatecheck,
24
- title = "{H}ate{C}heck: Functional Tests for Hate Speech Detection Models",
25
- author = {R{\"o}ttger, Paul and
26
- Vidgen, Bertie and
27
- Nguyen, Dong and
28
- Waseem, Zeerak and
29
- Margetts, Helen and
30
- Pierrehumbert, Janet},
31
- editor = "Zong, Chengqing and
32
- Xia, Fei and
33
- Li, Wenjie and
34
- Navigli, Roberto",
35
- 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)",
36
- month = aug,
37
- year = "2021",
38
- address = "Online",
39
- publisher = "Association for Computational Linguistics",
40
- url = "https://aclanthology.org/2021.acl-long.4",
41
- doi = "10.18653/v1/2021.acl-long.4",
42
- pages = "41--58",
43
- abstract = "Detecting online hate is a difficult task that even state-of-the-art models struggle with. Typically, hate speech detection models are evaluated by measuring their performance on held-out test data using metrics such as accuracy and F1 score. However, this approach makes it difficult to identify specific model weak points. It also risks overestimating generalisable model performance due to increasingly well-evidenced systematic gaps and biases in hate speech datasets. To enable more targeted diagnostic insights, we introduce HateCheck, a suite of functional tests for hate speech detection models. We specify 29 model functionalities motivated by a review of previous research and a series of interviews with civil society stakeholders. We craft test cases for each functionality and validate their quality through a structured annotation process. To illustrate HateCheck{'}s utility, we test near-state-of-the-art transformer models as well as two popular commercial models, revealing critical model weaknesses.",
44
- }
45
-
46
- @inproceedings{rottger-etal-2022-multilingual,
47
- title = "Multilingual {H}ate{C}heck: Functional Tests for Multilingual Hate Speech Detection Models",
48
- author = {R{\"o}ttger, Paul and
49
- Seelawi, Haitham and
50
- Nozza, Debora and
51
- Talat, Zeerak and
52
- Vidgen, Bertie},
53
- editor = "Narang, Kanika and
54
- Mostafazadeh Davani, Aida and
55
- Mathias, Lambert and
56
- Vidgen, Bertie and
57
- Talat, Zeerak",
58
- booktitle = "Proceedings of the Sixth Workshop on Online Abuse and Harms (WOAH)",
59
- month = jul,
60
- year = "2022",
61
- address = "Seattle, Washington (Hybrid)",
62
- publisher = "Association for Computational Linguistics",
63
- url = "https://aclanthology.org/2022.woah-1.15",
64
- doi = "10.18653/v1/2022.woah-1.15",
65
- pages = "154--169",
66
- abstract = "Hate speech detection models are typically evaluated on held-out test sets. However, this risks painting an incomplete and potentially misleading picture of model performance because of increasingly well-documented systematic gaps and biases in hate speech datasets. To enable more targeted diagnostic insights, recent research has thus introduced functional tests for hate speech detection models. However, these tests currently only exist for English-language content, which means that they cannot support the development of more effective models in other languages spoken by billions across the world. To help address this issue, we introduce Multilingual HateCheck (MHC), a suite of functional tests for multilingual hate speech detection models. MHC covers 34 functionalities across ten languages, which is more languages than any other hate speech dataset. To illustrate MHC{'}s utility, we train and test a high-performing multilingual hate speech detection model, and reveal critical model weaknesses for monolingual and cross-lingual applications.",
67
- }
68
- """
69
-
70
- _LICENSE = "Original datasets are released under cc-by-4.0."
71
-
72
-
73
- _LANGUAGES = {"ara": "Arabic",
74
- "cmn": "Mandarin",
75
- "eng": "English",
76
- "deu": "German",
77
- "fra": "French",
78
- "hin": "Hindi",
79
- "ita": "Italian",
80
- "nld": "Dutch",
81
- "pol": "Polish",
82
- "por": "Portuguese",
83
- "spa": "Spanish"}
84
-
85
- _ALL_LANGUAGES = "all_languages"
86
- _DOWNLOAD_URL = "{lang}/{split}.jsonl"
87
- _VERSION = "2.18.0"
88
-
89
-
90
- class MultiHatecheckConfig(datasets.BuilderConfig):
91
- """BuilderConfig for HatecheckMulti"""
92
-
93
- def __init__(self, languages=None, **kwargs):
94
- super(MultiHatecheckConfig, self).__init__(version=datasets.Version(_VERSION, ""), **kwargs),
95
- self.languages = languages
96
-
97
-
98
- class MultiHatecheck(datasets.GeneratorBasedBuilder):
99
-
100
- """Multilingual Hatecheck Corpus, by Paul Roettger"""
101
-
102
- BUILDER_CONFIGS = [
103
- MultiHatecheckConfig(
104
- name=_ALL_LANGUAGES,
105
- languages=_LANGUAGES,
106
- description="Hate speech detection dataset with binary (hateful vs non-hateful) labels. Includes 25+ distinct types of hate and challenging non-hate.",
107
- )
108
- ] + [
109
- MultiHatecheckConfig(
110
- name=lang,
111
- languages=[lang],
112
- description=f"{_LANGUAGES[lang]} examples of hate speech, with binary (hateful vs non-hateful) labels. Includes 25+ distinct types of hate and challenging non-hate.",
113
- )
114
- for lang in _LANGUAGES
115
- ]
116
- BUILDER_CONFIG_CLASS = MultiHatecheckConfig
117
- DEFAULT_CONFIG_NAME = _ALL_LANGUAGES
118
-
119
- def _info(self):
120
- return datasets.DatasetInfo(
121
- description=_DESCRIPTION,
122
- features=datasets.Features(
123
- {
124
- "text": datasets.Value("string"),
125
- "is_hateful": datasets.Value("string"),
126
- "functionality": datasets.Value("string"),
127
- }
128
- ),
129
- supervised_keys=None,
130
- license=_LICENSE,
131
- citation=_CITATION,
132
- )
133
-
134
- def _split_generators(self, dl_manager):
135
- test_urls = [_DOWNLOAD_URL.format(split="test", lang=lang) for lang in self.config.languages]
136
- test_paths = dl_manager.download_and_extract(test_urls)
137
-
138
- return [
139
- datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"file_paths": test_paths}),
140
- ]
141
-
142
- def _generate_examples(self, file_paths):
143
- row_count = 0
144
- for file_path in file_paths:
145
- with open(file_path, "r", encoding="utf-8") as f:
146
- for line in f:
147
- yield row_count, json.loads(line)
148
- row_count += 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
nld/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
pol/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
por/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
spa/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
test/ara.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4127f14a0ef09ecdae642ece021374d201aafbd32cf0322a408d06f0377c803
3
+ size 34084
test/cmn.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8b94e330b2d34a34649eb7cbb83bfd6eb18dde9e32c06a9263cd2ceebeafd3b
3
+ size 31171
test/deu.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97d79bcdf9b7619376277f5f08e0d88cedd8b1f9619a84004387664fa03d19ae
3
+ size 31764
test/eng.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a010ccc28fa52e1559699734ee4fef7352c1238fb9af65b2489866d12420afd0
3
+ size 32158
test/fra.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2160a864a5c1d690b6aec4ff53ae6fed352961edb532662885eb2985b38281c3
3
+ size 34431
test/hin.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dbbf57b1d15d0f1752ea2bd16e350ba23a8dccc91750a660254ade76af207ab
3
+ size 28928
test/ita.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1c592ccb362d4fb82d63cf43048b5288875dc18daa3aa82f8d9bc7bcb1e3ea9
3
+ size 31410
test/nld.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a163ee8e653ba62a5d65e8e7a56879133e2cc49cf3cb952820345c6bc5d78dc8
3
+ size 31667
test/pol.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2acdc0ade7187eed2a236ae030d41be0877d16dfb34b6ad95b00e3ce00874ac4
3
+ size 35639
test/por.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4481199b64d39c698b8874c9846c9eecaf2d7d035d1b4998dd06f634cedb4ea
3
+ size 31907
test/spa.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf749e41f48f63614cad5b88aeb0323bb2d800362c4a171ba6343109985b8d66
3
+ size 31592