Update README.md
Browse files
README.md
CHANGED
@@ -29,7 +29,9 @@ task_categories:
|
|
29 |
task_ids:
|
30 |
- conditional-text-generation-other-paraphrase-generation
|
31 |
---
|
32 |
-
|
|
|
|
|
33 |
## Table of Contents
|
34 |
- [Dataset Card Creation Guide](#dataset-card-creation-guide)
|
35 |
- [Table of Contents](#table-of-contents)
|
@@ -59,19 +61,27 @@ task_ids:
|
|
59 |
- [Licensing Information](#licensing-information)
|
60 |
- [Citation Information](#citation-information)
|
61 |
- [Contributions](#contributions)
|
|
|
62 |
## Dataset Description
|
|
|
63 |
- **Homepage:** https://indicnlp.ai4bharat.org/indicnlg-suite
|
64 |
- **Paper:** [IndicNLG Suite: Multilingual Datasets for Diverse NLG Tasks in Indic Languages](https://arxiv.org/abs/2203.05437)
|
65 |
- **Point of Contact:**
|
|
|
66 |
### Dataset Summary
|
|
|
67 |
IndicParaphrase is the paraphrasing dataset released as part of IndicNLG Suite. Each
|
68 |
input is paired with up to 5 references. We create this dataset in eleven
|
69 |
languages including as, bn, gu, hi, kn, ml, mr, or, pa, ta, te. The total
|
70 |
-
size of the dataset is 5.57M.
|
|
|
|
|
71 |
### Supported Tasks and Leaderboards
|
|
|
72 |
**Tasks:** Paraphrase generation
|
73 |
|
74 |
**Leaderboards:** Currently there is no Leaderboard for this dataset.
|
|
|
75 |
### Languages
|
76 |
- `Assamese (as)`
|
77 |
- `Bengali (bn)`
|
@@ -84,8 +94,11 @@ size of the dataset is 5.57M.
|
|
84 |
- `Punjabi (pa)`
|
85 |
- `Tamil (ta)`
|
86 |
- `Telugu (te)`
|
|
|
87 |
## Dataset Structure
|
|
|
88 |
### Data Instances
|
|
|
89 |
One example from the `hi` dataset is given below in JSON format.
|
90 |
```
|
91 |
{
|
@@ -99,15 +112,20 @@ One example from the `hi` dataset is given below in JSON format.
|
|
99 |
'target': 'प्रदेश के युवाओं को निजी उद्योगों में 75 प्रतिशत आरक्षण देंगे।'
|
100 |
}
|
101 |
```
|
|
|
102 |
### Data Fields
|
103 |
- `id (string)`: Unique identifier.
|
104 |
- `input (string)`: Input sentence
|
105 |
-
- `references (list of strings)`: Paraphrases of `input
|
106 |
-
- `target (string)`: The first reference (most dissimilar paraphrase)
|
|
|
107 |
|
108 |
### Data Splits
|
|
|
109 |
We first select 10K instances each for the validation and test and put remaining in the training dataset. `Assamese (as)`, due to its low-resource nature, could only be split into validation and test sets with 4,420 examples each.
|
110 |
Individual dataset with train-dev-test example counts are given below:
|
|
|
|
|
111 |
Language | ISO 639-1 Code |Train | Dev | Test |
|
112 |
--------------|----------------|-------|-----|------|
|
113 |
Assamese | as | - | 4,420 | 4,420 |
|
@@ -120,39 +138,76 @@ Marathi | mr |406,003 | 10,000 | 10,000 |
|
|
120 |
Oriya | or | 105,970 | 10,000 | 10,000 |
|
121 |
Punjabi | pa | 266,704 | 10,000 | 10,000 |
|
122 |
Tamil | ta | 497,798 | 10,000 | 10,000 |
|
123 |
-
Telugu | te | 596,283 | 10,000 | 10,000 |
|
|
|
124 |
|
125 |
## Dataset Creation
|
|
|
126 |
### Curation Rationale
|
|
|
127 |
[More information needed]
|
|
|
128 |
### Source Data
|
|
|
129 |
[Samanantar dataset](https://indicnlp.ai4bharat.org/samanantar/)
|
|
|
130 |
#### Initial Data Collection and Normalization
|
131 |
-
|
|
|
|
|
|
|
132 |
#### Who are the source language producers?
|
133 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
134 |
|
135 |
### Personal and Sensitive Information
|
|
|
136 |
[More information needed]
|
|
|
137 |
## Considerations for Using the Data
|
|
|
138 |
### Social Impact of Dataset
|
|
|
139 |
[More information needed]
|
|
|
140 |
### Discussion of Biases
|
|
|
141 |
[More information needed]
|
|
|
142 |
### Other Known Limitations
|
|
|
143 |
[More information needed]
|
|
|
144 |
## Additional Information
|
|
|
145 |
### Dataset Curators
|
|
|
146 |
[More information needed]
|
|
|
147 |
### Licensing Information
|
|
|
148 |
Contents of this repository are restricted to only non-commercial research purposes under the [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/). Copyright of the dataset contents belongs to the original copyright holders.
|
149 |
### Citation Information
|
|
|
150 |
If you use any of the datasets, models or code modules, please cite the following paper:
|
151 |
```
|
152 |
@inproceedings{Kumar2022IndicNLGSM,
|
153 |
title={IndicNLG Suite: Multilingual Datasets for Diverse NLG Tasks in Indic Languages},
|
154 |
author={Aman Kumar and Himani Shrotriya and Prachi Sahu and Raj Dabre and Ratish Puduppully and Anoop Kunchukuttan and Amogh Mishra and Mitesh M. Khapra and Pratyush Kumar},
|
155 |
year={2022},
|
156 |
-
url = "https://arxiv.org/abs/2203.05437"
|
157 |
```
|
|
|
|
|
158 |
### Contributions
|
|
|
|
29 |
task_ids:
|
30 |
- conditional-text-generation-other-paraphrase-generation
|
31 |
---
|
32 |
+
|
33 |
+
# Dataset Card for "IndicParaphrase"
|
34 |
+
|
35 |
## Table of Contents
|
36 |
- [Dataset Card Creation Guide](#dataset-card-creation-guide)
|
37 |
- [Table of Contents](#table-of-contents)
|
|
|
61 |
- [Licensing Information](#licensing-information)
|
62 |
- [Citation Information](#citation-information)
|
63 |
- [Contributions](#contributions)
|
64 |
+
|
65 |
## Dataset Description
|
66 |
+
|
67 |
- **Homepage:** https://indicnlp.ai4bharat.org/indicnlg-suite
|
68 |
- **Paper:** [IndicNLG Suite: Multilingual Datasets for Diverse NLG Tasks in Indic Languages](https://arxiv.org/abs/2203.05437)
|
69 |
- **Point of Contact:**
|
70 |
+
|
71 |
### Dataset Summary
|
72 |
+
|
73 |
IndicParaphrase is the paraphrasing dataset released as part of IndicNLG Suite. Each
|
74 |
input is paired with up to 5 references. We create this dataset in eleven
|
75 |
languages including as, bn, gu, hi, kn, ml, mr, or, pa, ta, te. The total
|
76 |
+
size of the dataset is 5.57M.
|
77 |
+
|
78 |
+
|
79 |
### Supported Tasks and Leaderboards
|
80 |
+
|
81 |
**Tasks:** Paraphrase generation
|
82 |
|
83 |
**Leaderboards:** Currently there is no Leaderboard for this dataset.
|
84 |
+
|
85 |
### Languages
|
86 |
- `Assamese (as)`
|
87 |
- `Bengali (bn)`
|
|
|
94 |
- `Punjabi (pa)`
|
95 |
- `Tamil (ta)`
|
96 |
- `Telugu (te)`
|
97 |
+
|
98 |
## Dataset Structure
|
99 |
+
|
100 |
### Data Instances
|
101 |
+
|
102 |
One example from the `hi` dataset is given below in JSON format.
|
103 |
```
|
104 |
{
|
|
|
112 |
'target': 'प्रदेश के युवाओं को निजी उद्योगों में 75 प्रतिशत आरक्षण देंगे।'
|
113 |
}
|
114 |
```
|
115 |
+
|
116 |
### Data Fields
|
117 |
- `id (string)`: Unique identifier.
|
118 |
- `input (string)`: Input sentence
|
119 |
+
- `references (list of strings)`: Paraphrases of `input`, ordered according to the least n-gram overlap
|
120 |
+
- `target (string)`: The first reference (most dissimilar paraphrase)
|
121 |
+
|
122 |
|
123 |
### Data Splits
|
124 |
+
|
125 |
We first select 10K instances each for the validation and test and put remaining in the training dataset. `Assamese (as)`, due to its low-resource nature, could only be split into validation and test sets with 4,420 examples each.
|
126 |
Individual dataset with train-dev-test example counts are given below:
|
127 |
+
|
128 |
+
|
129 |
Language | ISO 639-1 Code |Train | Dev | Test |
|
130 |
--------------|----------------|-------|-----|------|
|
131 |
Assamese | as | - | 4,420 | 4,420 |
|
|
|
138 |
Oriya | or | 105,970 | 10,000 | 10,000 |
|
139 |
Punjabi | pa | 266,704 | 10,000 | 10,000 |
|
140 |
Tamil | ta | 497,798 | 10,000 | 10,000 |
|
141 |
+
Telugu | te | 596,283 | 10,000 | 10,000 |
|
142 |
+
|
143 |
|
144 |
## Dataset Creation
|
145 |
+
|
146 |
### Curation Rationale
|
147 |
+
|
148 |
[More information needed]
|
149 |
+
|
150 |
### Source Data
|
151 |
+
|
152 |
[Samanantar dataset](https://indicnlp.ai4bharat.org/samanantar/)
|
153 |
+
|
154 |
#### Initial Data Collection and Normalization
|
155 |
+
|
156 |
+
[Detailed in the paper](https://arxiv.org/abs/2203.05437)
|
157 |
+
|
158 |
+
|
159 |
#### Who are the source language producers?
|
160 |
+
|
161 |
+
[Detailed in the paper](https://arxiv.org/abs/2203.05437)
|
162 |
+
|
163 |
+
|
164 |
+
### Annotations
|
165 |
+
[More information needed]
|
166 |
+
#### Annotation process
|
167 |
+
[More information needed]
|
168 |
+
|
169 |
+
#### Who are the annotators?
|
170 |
+
|
171 |
+
[More information needed]
|
172 |
|
173 |
### Personal and Sensitive Information
|
174 |
+
|
175 |
[More information needed]
|
176 |
+
|
177 |
## Considerations for Using the Data
|
178 |
+
|
179 |
### Social Impact of Dataset
|
180 |
+
|
181 |
[More information needed]
|
182 |
+
|
183 |
### Discussion of Biases
|
184 |
+
|
185 |
[More information needed]
|
186 |
+
|
187 |
### Other Known Limitations
|
188 |
+
|
189 |
[More information needed]
|
190 |
+
|
191 |
## Additional Information
|
192 |
+
|
193 |
### Dataset Curators
|
194 |
+
|
195 |
[More information needed]
|
196 |
+
|
197 |
### Licensing Information
|
198 |
+
|
199 |
Contents of this repository are restricted to only non-commercial research purposes under the [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/). Copyright of the dataset contents belongs to the original copyright holders.
|
200 |
### Citation Information
|
201 |
+
|
202 |
If you use any of the datasets, models or code modules, please cite the following paper:
|
203 |
```
|
204 |
@inproceedings{Kumar2022IndicNLGSM,
|
205 |
title={IndicNLG Suite: Multilingual Datasets for Diverse NLG Tasks in Indic Languages},
|
206 |
author={Aman Kumar and Himani Shrotriya and Prachi Sahu and Raj Dabre and Ratish Puduppully and Anoop Kunchukuttan and Amogh Mishra and Mitesh M. Khapra and Pratyush Kumar},
|
207 |
year={2022},
|
208 |
+
url = "https://arxiv.org/abs/2203.05437",
|
209 |
```
|
210 |
+
|
211 |
+
|
212 |
### Contributions
|
213 |
+
|