BSavoldi commited on
Commit
ba856a7
1 Parent(s): 0eddcb6

Upload 3 files

Browse files
Files changed (3) hide show
  1. GeNTE.tsv +0 -0
  2. GeNTE_common.tsv +0 -0
  3. README (1).md +172 -0
GeNTE.tsv ADDED
The diff for this file is too large to render. See raw diff
 
GeNTE_common.tsv ADDED
The diff for this file is too large to render. See raw diff
 
README (1).md ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ configs:
3
+ - config_name: main
4
+ data_files:
5
+ - split: test
6
+ path: "GeNTE.tsv"
7
+ default: true
8
+ - config_name: common
9
+ data_files:
10
+ - split: test
11
+ path: "GeNTE_common.tsv"
12
+
13
+
14
+
15
+ annotations_creators:
16
+ - expert-generated
17
+ language:
18
+ - en
19
+ - it
20
+ language_creators:
21
+ - expert-generated
22
+ license:
23
+ - cc-by-4.0
24
+ multilinguality:
25
+ - multilingual
26
+ - translation
27
+ paperswithcode_id: null
28
+ pretty_name: 'GeNTE: Gender-Neutral Translation Evaluation'
29
+ size_categories:
30
+ - 1K<n<10K
31
+ source_datasets: []
32
+ tags:
33
+ - gender
34
+ - bias
35
+ - inclusivity
36
+ - rewriting
37
+ - translation
38
+ - mt
39
+ task_categories:
40
+ - translation
41
+ - text-generation
42
+ task_ids:
43
+ - language-modeling
44
+ ---
45
+
46
+ # Dataset Card for GeNTE
47
+
48
+ **Homepage:** https://mt.fbk.eu/gente/
49
+
50
+
51
+ ### Dataset Summary
52
+
53
+ GeNTE (**Ge**nder-**N**eutral **T**ranslation **E**valuation) is a natural, bilingual corpus designed to benchmark the ability of machine translation systems to generate gender-neutral translations.
54
+
55
+ Built from European Parliament speeches, GeNTE comprises a subset of the English-Italian portion of the [Europarl corpus](https://www.statmt.org/europarl/archives.html).
56
+ GeNTE comprises 1500 parallel sentences, which are enriched with manual annotations and feature a balanced distribution of translation phenomena that either entail i) a gender-neutral translation (`set-N`), or ii) a gendered translation in the target language (`set-G`).
57
+
58
+
59
+
60
+ ### Supported Tasks and Languages
61
+
62
+ **Machine Translation**
63
+
64
+ GeNTE supports cross-lingual (en-it) and intra-lingual (it-it) gender inclusive translation tasks.
65
+
66
+ Refer to the paper [*Hi Guys* or *Hi Folks?* Benchmarking Gender-Neutral Machine Translation with the GeNTE Corpus](https://aclanthology.org/2023.emnlp-main.873/) for additional details on evaluation with GeNTE.
67
+
68
+ The evaluation code is available at [fbk-NEUTR-evAL](https://github.com/hlt-mt/fbk-NEUTR-evAL/blob/main/solutions/GeNTE.md).
69
+
70
+
71
+ ## Dataset Structure
72
+
73
+ ### Data Instances
74
+
75
+ The dataset consists of two configuration types (`main` and `common`) corrisponding to the files:
76
+
77
+ - **`GeNTE.tsv`:** The complete GeNTE corpus and its set annotations
78
+ - **`GeNTE_common.tsv`:** Subset of the GeNTE corpus that comprises 3 alternative gender-neutral reference translations
79
+
80
+
81
+ ### Data Fields
82
+
83
+ **`GeNTE.tsv`** is organized into 8 tab-separated columns as follows:
84
+
85
+ - ID: The unique GeNTE ID.
86
+ - Europarl_ID: The original sentence ID from Europarl's common-test-set 2.
87
+ - SET: Indicates whether the entry belongs to the Set-G or the Set-N subportion of the corpus.
88
+ - SRC: The English source sentence.
89
+ - REF-G: The gendered Italian reference translation.
90
+ - REF-N: The gender-neutral Italian reference, produced by a professional translator.
91
+ - COMMON: Indicates whether the entry is part of GeNTE common-set (yes/no).
92
+ - GENDER: For entries belonging to the Set-G, indicates if the the entry is Feminine or Masculine (F/M).
93
+
94
+ For entries of the common set, REF-N provides the gender-neutral Italian reference translation n. 2.
95
+
96
+ **`GeNTE-common.tsv`** comprises 200 entries organized into 9 tab-separated columns as follows:
97
+
98
+ - ID: The unique GeNTE ID.
99
+ - Europarl_ID: The original sentence ID from Europarl's common-test-set 2.
100
+ - SET: Indicates whether the entry belongs to the Set-G or the Set-N subportion of the corpus.
101
+ - SRC: The English source sentence.
102
+ - REF-G: The gendered Italian reference translation.
103
+ - REF-N1: The gender-neutral Italian reference produced by Translator 1.
104
+ - REF-N2: The gender-neutral Italian reference produced by Translator 2.
105
+ - REF-N3: The gender-neutral Italian reference produced by Translator 3.
106
+ - GENDER: For entries belonging to the Set-G, indicates if the the entry is Feminine or Masculine (F/M).
107
+
108
+
109
+ ## Dataset Creation
110
+
111
+ Refer to the original [paper](https://aclanthology.org/2023.emnlp-main.873/) for full details on dataset creation.
112
+
113
+ ### Curation Rationale
114
+
115
+ GeNTE is designed to evaluate models’ ability to perform gender-neutral translations under desirable circumstances. In
116
+ fact, when referents’ gender is unknown or irrelevant, undue gender inferences should not be made
117
+ and translation should be neutral. Instead, when a referent’s gender is relevant and
118
+ known, MT should not over-generalize to neutral translations. The corpus hence consists parallel sentences with mentions to human referents that equally represent two
119
+ translation scenarios:
120
+
121
+ - `Set-N`: featuring gender-ambiguous source sentences that require to be neutrally rendered in translation;
122
+ - `Set-G`: featuring gender-unambiguous source sentences, which shall
123
+ be properly rendered with gendered (masculine or feminine) forms in translation.
124
+
125
+ ### Source Data
126
+
127
+ The dataset contains text data extracted and edited from the Europarl Corpus ([common test set 2](https://www.statmt.org/europarl/archives.html)), and all rights of the data belong to the European Union and/or respective copyright holders.
128
+ Please refer to Europarl “[Terms of Use](https://www.statmt.org/europarl/archives.html)” for details.
129
+
130
+ ### Annotations
131
+ For each sentence pair extracted from Europarl (src, it-ref),GeNTE includes an additional Italian reference, which differs from the original one only in that it refers to
132
+ the human entities with neutral expressions.
133
+
134
+ The neutral reference translation were created by professionals based on the following [guidelines](https://drive.google.com/file/d/1TvV6NQoXiPHNSUHYlf4NFhef1_PKncF6/view?usp=sharing).
135
+
136
+ ### Dataset Curators
137
+
138
+ The authors of GeNTE are the dataset curators.
139
+
140
+ - Beatrice Savoldi (FBK): [email protected]
141
+ - Luisa Bentivogli (FBK): [email protected]
142
+ - Andrea Piergentili (FBK): [email protected]
143
+
144
+
145
+ ### Licensing Information
146
+
147
+ The GeNTE corpus is released under a Creative Commons Attribution 4.0 International license (CC BY 4.0).
148
+
149
+
150
+
151
+ ## Citation
152
+ ```bibtex
153
+ @inproceedings{piergentili-etal-2023-hi,
154
+ title = "Hi Guys or Hi Folks? Benchmarking Gender-Neutral Machine Translation with the {G}e{NTE} Corpus",
155
+ author = "Piergentili, Andrea and
156
+ Savoldi, Beatrice and
157
+ Fucci, Dennis and
158
+ Negri, Matteo and
159
+ Bentivogli, Luisa",
160
+ editor = "Bouamor, Houda and
161
+ Pino, Juan and
162
+ Bali, Kalika",
163
+ booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
164
+ month = dec,
165
+ year = "2023",
166
+ address = "Singapore",
167
+ publisher = "Association for Computational Linguistics",
168
+ url = "https://aclanthology.org/2023.emnlp-main.873",
169
+ doi = "10.18653/v1/2023.emnlp-main.873",
170
+ pages = "14124--14140"
171
+ }
172
+ ```