Update dataset card
Browse files
README.md
CHANGED
@@ -8,7 +8,7 @@ language:
|
|
8 |
- en
|
9 |
- yo
|
10 |
license:
|
11 |
-
- cc-by-4.0
|
12 |
multilinguality:
|
13 |
- translation
|
14 |
size_categories:
|
@@ -18,7 +18,7 @@ source_datasets:
|
|
18 |
task_categories:
|
19 |
- translation
|
20 |
task_ids: []
|
21 |
-
paperswithcode_id:
|
22 |
pretty_name: MENYO-20k
|
23 |
dataset_info:
|
24 |
features:
|
@@ -71,15 +71,15 @@ dataset_info:
|
|
71 |
|
72 |
## Dataset Description
|
73 |
|
74 |
-
- **Homepage:**
|
75 |
-
- **Repository:**
|
76 |
-
- **Paper:** [
|
77 |
- **Leaderboard:**
|
78 |
- **Point of Contact:**
|
79 |
|
80 |
### Dataset Summary
|
81 |
|
82 |
-
MENYO-20k is a multi-domain parallel dataset with texts obtained from news articles, ted talks, movie transcripts, radio transcripts, science and technology texts, and other short articles curated from the web and professional translators. The dataset has 20,100 parallel sentences split into 10,070 training sentences, 3,397 development sentences, and 6,633 test sentences (3,419 multi-domain, 1,714 news domain, and 1,500 ted talks speech transcript domain)
|
83 |
|
84 |
### Supported Tasks and Leaderboards
|
85 |
|
@@ -87,13 +87,13 @@ MENYO-20k is a multi-domain parallel dataset with texts obtained from news artic
|
|
87 |
|
88 |
### Languages
|
89 |
|
90 |
-
Languages are English and
|
91 |
|
92 |
## Dataset Structure
|
93 |
|
94 |
### Data Instances
|
95 |
|
96 |
-
|
97 |
|
98 |
```
|
99 |
{'translation':
|
@@ -101,18 +101,18 @@ The data consists of tab seperated entries
|
|
101 |
'yo': 'Ìdá 1: Kín ni Creative Commons?'
|
102 |
}
|
103 |
}
|
104 |
-
|
105 |
```
|
106 |
|
107 |
### Data Fields
|
108 |
|
109 |
-
- `
|
110 |
-
- `
|
|
|
111 |
|
112 |
|
113 |
### Data Splits
|
114 |
|
115 |
-
Training,
|
116 |
|
117 |
## Dataset Creation
|
118 |
|
@@ -166,25 +166,32 @@ Training, Validation and Test datasets are available.
|
|
166 |
|
167 |
### Licensing Information
|
168 |
|
169 |
-
The dataset is open but for non-commercial use because some
|
|
|
|
|
170 |
|
171 |
### Citation Information
|
|
|
|
|
172 |
```
|
173 |
-
@
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
-
|
|
|
|
|
|
|
188 |
}
|
189 |
```
|
190 |
### Contributions
|
|
|
8 |
- en
|
9 |
- yo
|
10 |
license:
|
11 |
+
- cc-by-nc-4.0
|
12 |
multilinguality:
|
13 |
- translation
|
14 |
size_categories:
|
|
|
18 |
task_categories:
|
19 |
- translation
|
20 |
task_ids: []
|
21 |
+
paperswithcode_id: menyo-20k
|
22 |
pretty_name: MENYO-20k
|
23 |
dataset_info:
|
24 |
features:
|
|
|
71 |
|
72 |
## Dataset Description
|
73 |
|
74 |
+
- **Homepage:**
|
75 |
+
- **Repository:** https://github.com/uds-lsv/menyo-20k_MT/
|
76 |
+
- **Paper:** [The Effect of Domain and Diacritics in Yorùbá-English Neural Machine Translation](https://arxiv.org/abs/2103.08647)
|
77 |
- **Leaderboard:**
|
78 |
- **Point of Contact:**
|
79 |
|
80 |
### Dataset Summary
|
81 |
|
82 |
+
MENYO-20k is a multi-domain parallel dataset with texts obtained from news articles, ted talks, movie transcripts, radio transcripts, science and technology texts, and other short articles curated from the web and professional translators. The dataset has 20,100 parallel sentences split into 10,070 training sentences, 3,397 development sentences, and 6,633 test sentences (3,419 multi-domain, 1,714 news domain, and 1,500 ted talks speech transcript domain).
|
83 |
|
84 |
### Supported Tasks and Leaderboards
|
85 |
|
|
|
87 |
|
88 |
### Languages
|
89 |
|
90 |
+
Languages are English and Yoruba.
|
91 |
|
92 |
## Dataset Structure
|
93 |
|
94 |
### Data Instances
|
95 |
|
96 |
+
An instance example:
|
97 |
|
98 |
```
|
99 |
{'translation':
|
|
|
101 |
'yo': 'Ìdá 1: Kín ni Creative Commons?'
|
102 |
}
|
103 |
}
|
|
|
104 |
```
|
105 |
|
106 |
### Data Fields
|
107 |
|
108 |
+
- `translation`:
|
109 |
+
- `en`: English sentence.
|
110 |
+
- `yo`: Yoruba sentence.
|
111 |
|
112 |
|
113 |
### Data Splits
|
114 |
|
115 |
+
Training, validation and test splits are available.
|
116 |
|
117 |
## Dataset Creation
|
118 |
|
|
|
166 |
|
167 |
### Licensing Information
|
168 |
|
169 |
+
The dataset is open but for non-commercial use because some data sources like Ted talks and JW news require permission for commercial use.
|
170 |
+
|
171 |
+
The dataset is licensed under Creative Commons [Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)](https://creativecommons.org/licenses/by-nc/4.0/) License: https://github.com/uds-lsv/menyo-20k_MT/blob/master/LICENSE
|
172 |
|
173 |
### Citation Information
|
174 |
+
|
175 |
+
If you use this dataset, please cite this paper:
|
176 |
```
|
177 |
+
@inproceedings{adelani-etal-2021-effect,
|
178 |
+
title = "The Effect of Domain and Diacritics in {Y}oruba{--}{E}nglish Neural Machine Translation",
|
179 |
+
author = "Adelani, David and
|
180 |
+
Ruiter, Dana and
|
181 |
+
Alabi, Jesujoba and
|
182 |
+
Adebonojo, Damilola and
|
183 |
+
Ayeni, Adesina and
|
184 |
+
Adeyemi, Mofe and
|
185 |
+
Awokoya, Ayodele Esther and
|
186 |
+
Espa{\~n}a-Bonet, Cristina",
|
187 |
+
booktitle = "Proceedings of the 18th Biennial Machine Translation Summit (Volume 1: Research Track)",
|
188 |
+
month = aug,
|
189 |
+
year = "2021",
|
190 |
+
address = "Virtual",
|
191 |
+
publisher = "Association for Machine Translation in the Americas",
|
192 |
+
url = "https://aclanthology.org/2021.mtsummit-research.6",
|
193 |
+
pages = "61--75",
|
194 |
+
abstract = "Massively multilingual machine translation (MT) has shown impressive capabilities and including zero and few-shot translation between low-resource language pairs. However and these models are often evaluated on high-resource languages with the assumption that they generalize to low-resource ones. The difficulty of evaluating MT models on low-resource pairs is often due to lack of standardized evaluation datasets. In this paper and we present MENYO-20k and the first multi-domain parallel corpus with a especially curated orthography for Yoruba{--}English with standardized train-test splits for benchmarking. We provide several neural MT benchmarks and compare them to the performance of popular pre-trained (massively multilingual) MT models both for the heterogeneous test set and its subdomains. Since these pre-trained models use huge amounts of data with uncertain quality and we also analyze the effect of diacritics and a major characteristic of Yoruba and in the training data. We investigate how and when this training condition affects the final quality of a translation and its understandability.Our models outperform massively multilingual models such as Google ($+8.7$ BLEU) and Facebook M2M ($+9.1$) when translating to Yoruba and setting a high quality benchmark for future research.",
|
195 |
}
|
196 |
```
|
197 |
### Contributions
|