asi commited on
Commit
033d2bd
1 Parent(s): 62d39da

:books: add first dataset documentation

Browse files
Files changed (1) hide show
  1. README.md +54 -95
README.md CHANGED
@@ -1,6 +1,23 @@
1
  ---
2
- YAML tags:
3
- - copy-paste the tags obtained with the tagging app: https://github.com/huggingface/datasets-tagging
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
  ---
5
 
6
  # Dataset Card Creation Guide
@@ -37,165 +54,107 @@ YAML tags:
37
 
38
  ## Dataset Description
39
 
40
- - **Homepage:** [Add homepage URL here if available (unless it's a GitHub repository)]()
41
- - **Repository:** [If the dataset is hosted on github or has a github homepage, add URL here]()
42
- - **Paper:** [If the dataset was introduced by a paper or there was a paper written describing the dataset, add URL here (landing page for Arxiv paper preferred)]()
43
- - **Leaderboard:** [If the dataset supports an active leaderboard, add link here]()
44
- - **Point of Contact:** [If known, name and email of at least one person the reader can contact for questions about the dataset.]()
45
 
46
  ### Dataset Summary
47
 
48
- Wikitext-fr language modeling dataset consists of over 70 million tokens extracted from the set of french Wikipedia articles that are classified as "quality articles" or "good articles.". It is designed to mirror the english benchmark from Stephen Merity, Caiming Xiong, James Bradbury, and Richard Socher. 2016.
49
  [Pointer Sentinel Mixture Models](https://arxiv.org/abs/1609.07843) The dataset is available under the [Creative Commons Attribution-ShareAlike License](https://creativecommons.org/licenses/by-sa/4.0/)
50
 
51
  ### Supported Tasks and Leaderboards
52
 
53
- For each of the tasks tagged for this dataset, give a brief description of the tag, metrics, and suggested models (with a link to their HuggingFace implementation if available). Give a similar description of tasks that were not covered by the structured tag set (repace the `task-category-tag` with an appropriate `other:other-task-name`).
54
-
55
- - `task-category-tag`: The dataset can be used to train a model for [TASK NAME], which consists in [TASK DESCRIPTION]. Success on this task is typically measured by achieving a *high/low* [metric name](https://huggingface.co/metrics/metric_name). The ([model name](https://huggingface.co/model_name) or [model class](https://huggingface.co/transformers/model_doc/model_class.html)) model currently achieves the following score. *[IF A LEADERBOARD IS AVAILABLE]:* This task has an active leaderboard which can be found at [leaderboard url]() and ranks models based on [metric name](https://huggingface.co/metrics/metric_name) while also reporting [other metric name](https://huggingface.co/metrics/other_metric_name).
56
 
57
  ### Languages
58
 
59
- Provide a brief overview of the languages represented in the dataset. Describe relevant details about specifics of the language such as whether it is social media text, African American English,...
60
-
61
- When relevant, please provide [BCP-47 codes](https://tools.ietf.org/html/bcp47), which consist of a [primary language subtag](https://tools.ietf.org/html/bcp47#section-2.2.1), with a [script subtag](https://tools.ietf.org/html/bcp47#section-2.2.3) and/or [region subtag](https://tools.ietf.org/html/bcp47#section-2.2.4) if available.
62
 
63
  ## Dataset Structure
64
 
65
  ### Data Instances
66
 
67
- Provide an JSON-formatted example and brief description of a typical instance in the dataset. If available, provide a link to further examples.
68
 
69
  ```
70
  {
71
- 'example_field': ...,
72
  ...
73
  }
74
  ```
75
 
76
- Provide any additional information that is not covered in the other sections about the data here. In particular describe any relationships between data points and if these relationships are made explicit.
77
 
78
  ### Data Fields
79
 
80
- List and describe the fields present in the dataset. Mention their data type, and whether they are used as input or output in any of the tasks the dataset currently supports. If the data has span indices, describe their attributes, such as whether they are at the character level or word level, whether they are contiguous or not, etc. If the datasets contains example IDs, state whether they have an inherent meaning, such as a mapping to other datasets or pointing to relationships between data points.
81
-
82
- - `example_field`: description of `example_field`
83
-
84
- Note that the descriptions can be initialized with the **Show Markdown Data Fields** output of the [tagging app](https://github.com/huggingface/datasets-tagging), you will then only need to refine the generated descriptions.
85
 
86
  ### Data Splits
87
 
88
- Describe and name the splits in the dataset if there are more than one.
89
 
90
- Describe any criteria for splitting the data, if used. If their are differences between the splits (e.g. if the training annotations are machine-generated and the dev and test ones are created by humans, or if different numbers of annotators contributed to each example), describe them here.
 
 
 
 
 
91
 
92
- Provide the sizes of each split. As appropriate, provide any descriptive statistics for the features, such as average length. For example:
93
-
94
- | | Tain | Valid | Test |
95
- | ----- | ------ | ----- | ---- |
96
- | Input Sentences | | | |
97
- | Average Sentence Length | | | |
98
 
99
  ## Dataset Creation
100
 
101
  ### Curation Rationale
102
 
103
- What need motivated the creation of this dataset? What are some of the reasons underlying the major choices involved in putting it together?
104
 
105
  ### Source Data
106
 
107
- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences,...)
 
108
 
109
  #### Initial Data Collection and Normalization
110
 
111
- Describe the data collection process. Describe any criteria for data selection or filtering. List any key words or search terms used. If possible, include runtime information for the collection process.
112
-
113
- If data was collected from other pre-existing datasets, link to source here and to their [Hugging Face version](https://huggingface.co/datasets/dataset_name).
114
-
115
- If the data was modified or normalized after being collected (e.g. if the data is word-tokenized), describe the process and the tools used.
116
-
117
- #### Who are the source language producers?
118
-
119
- State whether the data was produced by humans or machine generated. Describe the people or systems who originally created the data.
120
-
121
- If available, include self-reported demographic or identity information for the source data creators, but avoid inferring this information. Instead state that this information is unknown. See [Larson 2017](https://www.aclweb.org/anthology/W17-1601.pdf) for using identity categories as a variables, particularly gender.
122
-
123
- Describe the conditions under which the data was created (for example, if the producers were crowdworkers, state what platform was used, or if the data was found, what website the data was found on). If compensation was provided, include that information here.
124
-
125
- Describe other people represented or mentioned in the data. Where possible, link to references for the information.
126
-
127
- ### Annotations
128
-
129
- If the dataset contains annotations which are not part of the initial data collection, describe them in the following paragraphs.
130
-
131
- #### Annotation process
132
-
133
- If applicable, describe the annotation process and any tools used, or state otherwise. Describe the amount of data annotated, if not all. Describe or reference annotation guidelines provided to the annotators. If available, provide interannotator statistics. Describe any annotation validation processes.
134
 
135
- #### Who are the annotators?
136
-
137
- If annotations were collected for the source data (such as class labels or syntactic parses), state whether the annotations were produced by humans or machine generated.
138
-
139
- Describe the people or systems who originally created the annotations and their selection criteria if applicable.
140
-
141
- If available, include self-reported demographic or identity information for the annotators, but avoid inferring this information. Instead state that this information is unknown. See [Larson 2017](https://www.aclweb.org/anthology/W17-1601.pdf) for using identity categories as a variables, particularly gender.
142
-
143
- Describe the conditions under which the data was annotated (for example, if the annotators were crowdworkers, state what platform was used, or if the data was found, what website the data was found on). If compensation was provided, include that information here.
144
 
145
  ### Personal and Sensitive Information
146
 
147
- State whether the dataset uses identity categories and, if so, how the information is used. Describe where this information comes from (i.e. self-reporting, collecting from profiles, inferring, etc.). See [Larson 2017](https://www.aclweb.org/anthology/W17-1601.pdf) for using identity categories as a variables, particularly gender. State whether the data is linked to individuals and whether those individuals can be identified in the dataset, either directly or indirectly (i.e., in combination with other data).
148
-
149
- State whether the dataset contains other data that might be considered sensitive (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history).
150
-
151
- If efforts were made to anonymize the data, describe the anonymization process.
152
-
153
  ## Considerations for Using the Data
154
 
155
  ### Social Impact of Dataset
156
 
157
- Please discuss some of the ways you believe the use of this dataset will impact society.
158
-
159
- The statement should include both positive outlooks, such as outlining how technologies developed through its use may improve people's lives, and discuss the accompanying risks. These risks may range from making important decisions more opaque to people who are affected by the technology, to reinforcing existing harmful biases (whose specifics should be discussed in the next section), among other considerations.
160
-
161
- Also describe in this section if the proposed dataset contains a low-resource or under-represented language. If this is the case or if this task has any impact on underserved communities, please elaborate here.
162
-
163
  ### Discussion of Biases
164
 
165
- Provide descriptions of specific biases that are likely to be reflected in the data, and state whether any steps were taken to reduce their impact.
166
-
167
- For Wikipedia text, see for example [Dinan et al 2020 on biases in Wikipedia (esp. Table 1)](https://arxiv.org/abs/2005.00614), or [Blodgett et al 2020](https://www.aclweb.org/anthology/2020.acl-main.485/) for a more general discussion of the topic.
168
-
169
- If analyses have been run quantifying these biases, please add brief summaries and links to the studies here.
170
-
171
  ### Other Known Limitations
172
 
173
- If studies of the datasets have outlined other limitations of the dataset, such as annotation artifacts, please outline and cite them here.
174
-
175
  ## Additional Information
176
 
177
  ### Dataset Curators
178
 
179
- List the people involved in collecting the dataset and their affiliation(s). If funding information is known, include it here.
180
-
181
  ### Licensing Information
182
 
183
- Provide the license and link to the license webpage if available.
184
 
185
  ### Citation Information
186
 
187
- Provide the [BibTex](http://www.bibtex.org/)-formatted reference for the dataset. For example:
188
  ```
189
- @article{article_id,
190
- author = {Author List},
191
- title = {Dataset Paper Title},
192
- journal = {Publication Venue},
193
- year = {2525}
 
 
 
 
 
 
 
 
 
194
  }
195
  ```
196
 
197
- If the dataset has a [DOI](https://www.doi.org/), please provide it here.
198
 
199
  ### Contributions
200
 
201
- Thanks to [@github-username](https://github.com/<github-username>) for adding this dataset.
1
  ---
2
+ annotations_creators:
3
+ - no-annotation
4
+ language_creators:
5
+ - found
6
+ languages:
7
+ - fr-FR
8
+ licenses:
9
+ - cc-by-sa-4.0
10
+ multilinguality:
11
+ - monolingual
12
+ pretty_name: Wikitext-fr
13
+ size_categories:
14
+ - unknown
15
+ source_datasets:
16
+ - original
17
+ task_categories:
18
+ - sequence-modeling
19
+ task_ids:
20
+ - language-modeling
21
  ---
22
 
23
  # Dataset Card Creation Guide
54
 
55
  ## Dataset Description
56
 
57
+ - **Repository:** [https://github.com/AntoineSimoulin/gpt-fr](https://github.com/AntoineSimoulin/gpt-fr)
58
+ - **Paper:** [https://aclanthology.org/2021.jeptalnrecital-taln.24.pdf](https://aclanthology.org/2021.jeptalnrecital-taln.24.pdf)
 
 
 
59
 
60
  ### Dataset Summary
61
 
62
+ Wikitext-fr language modeling dataset consists of over 70 million tokens extracted from the set of french Wikipedia articles that are classified as "quality articles" or "good articles". It is designed to mirror the english benchmark from Stephen Merity, Caiming Xiong, James Bradbury, and Richard Socher. 2016.
63
  [Pointer Sentinel Mixture Models](https://arxiv.org/abs/1609.07843) The dataset is available under the [Creative Commons Attribution-ShareAlike License](https://creativecommons.org/licenses/by-sa/4.0/)
64
 
65
  ### Supported Tasks and Leaderboards
66
 
67
+ - `language-modeling`: The dataset can be used to evaluate the generation abilites of a model. Success on this task is typically measured by achieving a *low* perplexity. The ([model name](https://huggingface.co/asi/gpt-fr-cased-base) currently achieves 12.9.
 
 
68
 
69
  ### Languages
70
 
71
+ The dataset is in French.
 
 
72
 
73
  ## Dataset Structure
74
 
75
  ### Data Instances
76
 
77
+ The dataset consists in the agregation of paragraphs from wikipedia articles.
78
 
79
  ```
80
  {
81
+ 'paragraph': ...,
82
  ...
83
  }
84
  ```
85
 
 
86
 
87
  ### Data Fields
88
 
89
+ - `paragraph`: This is a paragraph from the original wikipedia article.
 
 
 
 
90
 
91
  ### Data Splits
92
 
93
+ The dataset is splited into a train/valid/test split.
94
 
95
+ | | Tain (35) | Train (72) | Valid | Test |
96
+ | ----- | ------ | ----- | ---- | ---- |
97
+ | Number of Documents | 2 126 | 5 902 | 60 | 60 |
98
+ | Number of tokens | 351 66 | 72 961 | 896 | 897 |
99
+ | Vocabulary size | 137 589 | 205 403 | | |
100
+ | Out of Vocabulary | 0.8% | 1.2% | | |
101
 
 
 
 
 
 
 
102
 
103
  ## Dataset Creation
104
 
105
  ### Curation Rationale
106
 
107
+ The dataset is created to evaluate French models with similart criteria than English.s
108
 
109
  ### Source Data
110
 
111
+ Wikitext-fr language modeling dataset consists of over 70 million tokens extracted from the set of french Wikipedia articles that are classified as "quality articles" or "good articles".
112
+ We did not apply specific pre-treatments as transformers models might use a dedicated tokenization.s
113
 
114
  #### Initial Data Collection and Normalization
115
 
116
+ We used the Wikipedia API to collect the articles since cleaning Wikipedia articles from dumps is not a trivial task.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
117
 
 
 
 
 
 
 
 
 
 
118
 
119
  ### Personal and Sensitive Information
120
 
 
 
 
 
 
 
121
  ## Considerations for Using the Data
122
 
123
  ### Social Impact of Dataset
124
 
 
 
 
 
 
 
125
  ### Discussion of Biases
126
 
 
 
 
 
 
 
127
  ### Other Known Limitations
128
 
 
 
129
  ## Additional Information
130
 
131
  ### Dataset Curators
132
 
 
 
133
  ### Licensing Information
134
 
135
+ The dataset is available under the [Creative Commons Attribution-ShareAlike License](https://creativecommons.org/licenses/by-sa/4.0/)
136
 
137
  ### Citation Information
138
 
 
139
  ```
140
+ @inproceedings{simoulin:hal-03265900,
141
+ TITLE = {{Un mod{\`e}le Transformer G{\'e}n{\'e}ratif Pr{\'e}-entrain{\'e} pour le \_\_\_\_\_\_ fran{\c c}ais}},
142
+ AUTHOR = {Simoulin, Antoine and Crabb{\'e}, Benoit},
143
+ URL = {https://hal.archives-ouvertes.fr/hal-03265900},
144
+ BOOKTITLE = {{Traitement Automatique des Langues Naturelles}},
145
+ ADDRESS = {Lille, France},
146
+ EDITOR = {Denis, Pascal and Grabar, Natalia and Fraisse, Amel and Cardon, R{\'e}mi and Jacquemin, Bernard and Kergosien, Eric and Balvet, Antonio},
147
+ PUBLISHER = {{ATALA}},
148
+ PAGES = {246-255},
149
+ YEAR = {2021},
150
+ KEYWORDS = {fran{\c c}ais. ; GPT ; G{\'e}n{\'e}ratif ; Transformer ; Pr{\'e}-entra{\^i}n{\'e}},
151
+ PDF = {https://hal.archives-ouvertes.fr/hal-03265900/file/7.pdf},
152
+ HAL_ID = {hal-03265900},
153
+ HAL_VERSION = {v1},
154
  }
155
  ```
156
 
 
157
 
158
  ### Contributions
159
 
160
+ Thanks to [@AntoineSimoulin](https://github.com/AntoineSimoulin) for adding this dataset.