Datasets:
Added dataset card
Browse files- README.md +157 -1
- mt_geneval.py +4 -4
README.md
CHANGED
@@ -1,3 +1,159 @@
|
|
1 |
---
|
2 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
annotations_creators:
|
3 |
+
- expert-generated
|
4 |
+
language:
|
5 |
+
- en
|
6 |
+
- it
|
7 |
+
- fr
|
8 |
+
- ar
|
9 |
+
- de
|
10 |
+
- hi
|
11 |
+
- pt
|
12 |
+
- ru
|
13 |
+
- es
|
14 |
+
language_creators:
|
15 |
+
- expert-generated
|
16 |
+
license:
|
17 |
+
- cc-by-sa-3.0
|
18 |
+
multilinguality:
|
19 |
+
- translation
|
20 |
+
pretty_name: mt_geneval
|
21 |
+
size_categories:
|
22 |
+
- 1K<n<10K
|
23 |
+
source_datasets:
|
24 |
+
- original
|
25 |
+
tags:
|
26 |
+
- gender
|
27 |
+
- constrained mt
|
28 |
+
task_categories:
|
29 |
+
- translation
|
30 |
+
task_ids: []
|
31 |
---
|
32 |
+
|
33 |
+
# Dataset Card for MT-GenEval
|
34 |
+
|
35 |
+
## Table of Contents
|
36 |
+
|
37 |
+
- [Dataset Card for MT-GenEval](#dataset-card-for-mt-geneval)
|
38 |
+
- [Table of Contents](#table-of-contents)
|
39 |
+
- [Dataset Description](#dataset-description)
|
40 |
+
- [Dataset Summary](#dataset-summary)
|
41 |
+
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
|
42 |
+
- [Machine Translation](#machine-translation)
|
43 |
+
- [Languages](#languages)
|
44 |
+
- [Dataset Structure](#dataset-structure)
|
45 |
+
- [Data Instances](#data-instances)
|
46 |
+
- [Data Splits](#data-splits)
|
47 |
+
- [Dataset Creation](#dataset-creation)
|
48 |
+
- [Additional Information](#additional-information)
|
49 |
+
- [Dataset Curators](#dataset-curators)
|
50 |
+
- [Licensing Information](#licensing-information)
|
51 |
+
- [Citation Information](#citation-information)
|
52 |
+
|
53 |
+
## Dataset Description
|
54 |
+
|
55 |
+
- **Repository:** [Github](https://github.com/amazon-science/machine-translation-gender-eval)
|
56 |
+
- **Paper:** [EMNLP 2022](https://arxiv.org/abs/2211.01355)
|
57 |
+
- **Point of Contact:** [Anna Currey](mailto:[email protected])
|
58 |
+
|
59 |
+
### Dataset Summary
|
60 |
+
|
61 |
+
The MT-GenEval benchmark evaluates gender translation accuracy on English -> {Arabic, French, German, Hindi, Italian, Portuguese, Russian, Spanish}. The dataset contains individual sentences with annotations on the gendered target words, and contrastive original-invertend translations with additional preceding context.
|
62 |
+
|
63 |
+
**Disclaimer**: *The MT-GenEval benchmark was released in the EMNLP 2022 paper [MT-GenEval: A Counterfactual and Contextual Dataset for Evaluating Gender Accuracy in Machine Translation](https://arxiv.org/abs/2211.01355) by Anna Currey, Maria Nadejde, Raghavendra Pappagari, Mia Mayer, Stanislas Lauly, Xing Niu, Benjamin Hsu, and Georgiana Dinu and is hosted through Github by the [Amazon Science](https://github.com/amazon-science?type=source) organization. The dataset is licensed under a [Creative Commons Attribution-ShareAlike 3.0 Unported License](https://creativecommons.org/licenses/by-sa/3.0/).*
|
64 |
+
|
65 |
+
### Supported Tasks and Leaderboards
|
66 |
+
#### Machine Translation
|
67 |
+
Refer to the [original paper](https://arxiv.org/abs/2211.01355) for additional details on gender accuracy evaluation with MT-GenEval.
|
68 |
+
### Languages
|
69 |
+
The dataset contains source English sentences extracted from Wikipedia translated into the following languages: Arabic (`ar`), French (`fr`), German (`de`), Hindi (`hi`), Italian (`it`), Portuguese (`pt`), Russian (`ru`), and Spanish (`es`).
|
70 |
+
## Dataset Structure
|
71 |
+
### Data Instances
|
72 |
+
|
73 |
+
The dataset contains two configuration types, `sentences` and `context`, mirroring the original repository structure, with source and target language specified in the configuration name (e.g. `sentences_en_ar`, `context_en_it`) The `sentences` configurations contains masculine and feminine versions of individual sentences with gendered word annotations. Here is an example entry of the `sentences_en_it` split (all `sentences_en_XX` splits have the same structure):
|
74 |
+
|
75 |
+
```json
|
76 |
+
{
|
77 |
+
{
|
78 |
+
"orig_id": 0,
|
79 |
+
"source_feminine": "Pagratidis quickly recanted her confession, claiming she was psychologically pressured and beaten, and until the moment of her execution, she remained firm in her innocence.",
|
80 |
+
"reference_feminine": "Pagratidis subito ritrattò la sua confessione, affermando che era aveva subito pressioni psicologiche e era stata picchiata, e fino al momento della sua esecuzione, rimase ferma sulla sua innocenza.",
|
81 |
+
"source_masculine": "Pagratidis quickly recanted his confession, claiming he was psychologically pressured and beaten, and until the moment of his execution, he remained firm in his innocence.",
|
82 |
+
"reference_masculine": "Pagratidis subito ritrattò la sua confessione, affermando che era aveva subito pressioni psicologiche e era stato picchiato, e fino al momento della sua esecuzione, rimase fermo sulla sua innocenza.",
|
83 |
+
"source_feminine_annotated": "Pagratidis quickly recanted <F>her</F> confession, claiming <F>she</F> was psychologically pressured and beaten, and until the moment of <F>her</F> execution, <F>she</F> remained firm in <F>her</F> innocence.",
|
84 |
+
"reference_feminine_annotated": "Pagratidis subito ritrattò la sua confessione, affermando che era aveva subito pressioni psicologiche e era <F>stata picchiata</F>, e fino al momento della sua esecuzione, rimase <F>ferma</F> sulla sua innocenza.",
|
85 |
+
"source_masculine_annotated": "Pagratidis quickly recanted <M>his</M> confession, claiming <M>he</M> was psychologically pressured and beaten, and until the moment of <M>his</M> execution, <M>he</M> remained firm in <M>his</M> innocence.",
|
86 |
+
"reference_masculine_annotated": "Pagratidis subito ritrattò la sua confessione, affermando che era aveva subito pressioni psicologiche e era <M>stato picchiato</M>, e fino al momento della sua esecuzione, rimase <M>fermo</M> sulla sua innocenza.",
|
87 |
+
"source_feminine_keywords": "her;she;her;she;her",
|
88 |
+
"reference_feminine_keywords": "stata picchiata;ferma",
|
89 |
+
"source_masculine_keywords": "his;he;his;he;his",
|
90 |
+
"reference_masculine_keywords": "stato picchiato;fermo",
|
91 |
+
}
|
92 |
+
}
|
93 |
+
```
|
94 |
+
|
95 |
+
The `context` configuration contains instead different English sources related to stereotypical professional roles with additional preceding context and contrastive original-inverted translations. Here is an example entry of the `context_en_it` split (all `context_en_XX` splits have the same structure):
|
96 |
+
|
97 |
+
```json
|
98 |
+
{
|
99 |
+
"orig_id": 0,
|
100 |
+
"context": "Pierpont told of entering and holding up the bank and then fleeing to Fort Wayne, where the loot was divided between him and three others.",
|
101 |
+
"source": "However, Pierpont stated that Skeer was the planner of the robbery.",
|
102 |
+
"reference_original": "Comunque, Pierpont disse che Skeer era il pianificatore della rapina.",
|
103 |
+
"reference_flipped": "Comunque, Pierpont disse che Skeer era la pianificatrice della rapina."
|
104 |
+
}
|
105 |
+
```
|
106 |
+
|
107 |
+
### Data Splits
|
108 |
+
|
109 |
+
All `sentences_en_XX` configurations have 1200 examples in the `train` split and 300 in the `test` split. For the `context_en_XX` configurations, the number of example depends on the language pair:
|
110 |
+
|
111 |
+
| Configuration | # Train | # Test |
|
112 |
+
| :-----------: | :--------: | :-----: |
|
113 |
+
| `context_en_ar` | 792 | 1100 |
|
114 |
+
| `context_en_fr` | 477 | 1099 |
|
115 |
+
| `context_en_de` | 598 | 1100 |
|
116 |
+
| `context_en_hi` | 397 | 1098 |
|
117 |
+
| `context_en_it` | 465 | 1904 |
|
118 |
+
| `context_en_pt` | 574 | 1089 |
|
119 |
+
| `context_en_ru` | 583 | 1100 |
|
120 |
+
| `context_en_es` | 534 | 1096 |
|
121 |
+
|
122 |
+
### Dataset Creation
|
123 |
+
|
124 |
+
From the original paper:
|
125 |
+
|
126 |
+
>In developing MT-GenEval, our goal was to create a realistic, gender-balanced dataset that naturally incorporates a diverse range of gender phenomena. To this end, we extracted English source sentences from Wikipedia as the basis for our dataset. We automatically pre-selected relevant sentences using EN gender-referring words based on the list provided by [Zhao et al. (2018)](https://doi.org/10.18653/v1/N18-2003).
|
127 |
+
|
128 |
+
Please refer to the original article [MT-GenEval: A Counterfactual and Contextual Dataset for Evaluating Gender Accuracy in Machine Translation](https://arxiv.org/abs/2211.01355) for additional information on dataset creation.
|
129 |
+
|
130 |
+
## Additional Information
|
131 |
+
### Dataset Curators
|
132 |
+
|
133 |
+
The original authors of MT-GenEval are the curators of the original dataset. For problems or updates on this 🤗 Datasets version, please contact [[email protected]](mailto:[email protected]).
|
134 |
+
|
135 |
+
### Licensing Information
|
136 |
+
|
137 |
+
The dataset is licensed under the [Creative Commons Attribution-ShareAlike 3.0 International License](https://creativecommons.org/licenses/by-sa/3.0/).
|
138 |
+
|
139 |
+
### Citation Information
|
140 |
+
Please cite the authors if you use these corpora in your work.
|
141 |
+
|
142 |
+
```bibtex
|
143 |
+
@inproceedings{currey-etal-2022-mtgeneval,
|
144 |
+
title = "{MT-GenEval}: {A} Counterfactual and Contextual Dataset for Evaluating Gender Accuracy in Machine Translation",
|
145 |
+
author = "Currey, Anna and
|
146 |
+
Nadejde, Maria and
|
147 |
+
Pappagari, Raghavendra and
|
148 |
+
Mayer, Mia and
|
149 |
+
Lauly, Stanislas, and
|
150 |
+
Niu, Xing and
|
151 |
+
Hsu, Benjamin and
|
152 |
+
Dinu, Georgiana",
|
153 |
+
booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
|
154 |
+
month = dec,
|
155 |
+
year = "2022",
|
156 |
+
publisher = "Association for Computational Linguistics",
|
157 |
+
url = "https://arxiv.org/abs/2211.01355",
|
158 |
+
}
|
159 |
+
```
|
mt_geneval.py
CHANGED
@@ -211,10 +211,10 @@ class WmtVat(datasets.GeneratorBasedBuilder):
|
|
211 |
"reference_feminine_annotated": rfa,
|
212 |
"source_masculine_annotated": sma,
|
213 |
"reference_masculine_annotated": rma,
|
214 |
-
"source_feminine_keywords": sfk,
|
215 |
-
"reference_feminine_keywords": rfk,
|
216 |
-
"source_masculine_keywords": smk,
|
217 |
-
"reference_masculine_keywords": rmk
|
218 |
}
|
219 |
else:
|
220 |
with open(filepaths["2to1"]) as f:
|
|
|
211 |
"reference_feminine_annotated": rfa,
|
212 |
"source_masculine_annotated": sma,
|
213 |
"reference_masculine_annotated": rma,
|
214 |
+
"source_feminine_keywords": ";".join(sfk),
|
215 |
+
"reference_feminine_keywords": ";".join(rfk),
|
216 |
+
"source_masculine_keywords": ";".join(smk),
|
217 |
+
"reference_masculine_keywords": ";".join(rmk)
|
218 |
}
|
219 |
else:
|
220 |
with open(filepaths["2to1"]) as f:
|