oooooorange
commited on
Commit
•
c1db91a
1
Parent(s):
411068d
initial commit
Browse files
README.md
CHANGED
@@ -21,4 +21,95 @@ language:
|
|
21 |
pretty_name: varta
|
22 |
size_categories:
|
23 |
- 10M<n<100M
|
24 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
pretty_name: varta
|
22 |
size_categories:
|
23 |
- 10M<n<100M
|
24 |
+
---
|
25 |
+
# Dataset Card for Varta
|
26 |
+
|
27 |
+
## Dataset Description
|
28 |
+
|
29 |
+
- **Homepage:**
|
30 |
+
- **Repository:**
|
31 |
+
- **Paper:**
|
32 |
+
- **Leaderboard:**
|
33 |
+
- **Point of Contact:**
|
34 |
+
|
35 |
+
### Dataset Summary
|
36 |
+
|
37 |
+
Varta is a diverse, challenging, large-scale, multilingual, high-quality headline generation dataset containing 41.8 million news articles in 14 Indic languages and English.
|
38 |
+
The data is crawled from DailyHunt, a popular news aggregator in India that pulls high-quality articles from multiple trusted and reputed news publishers.
|
39 |
+
|
40 |
+
### Languages
|
41 |
+
|
42 |
+
- as
|
43 |
+
- bh
|
44 |
+
- bn
|
45 |
+
- en
|
46 |
+
- gu
|
47 |
+
- hi
|
48 |
+
- kn
|
49 |
+
- ml
|
50 |
+
- mr
|
51 |
+
- ne
|
52 |
+
- or
|
53 |
+
- pa
|
54 |
+
- ta
|
55 |
+
- te
|
56 |
+
- ur
|
57 |
+
|
58 |
+
## Dataset Structure
|
59 |
+
|
60 |
+
### Data Instances
|
61 |
+
```
|
62 |
+
{
|
63 |
+
"id":"n400000150",
|
64 |
+
"langCode":"as",
|
65 |
+
"source_url":"https://www.etvbharat.com/assamese/assam/bharat/militant-hideout-destroyed-on-srinagar-bandipora-highway/assam20220630074145729729173",
|
66 |
+
"dh_url":"https://m.dailyhunt.in/news/india/assamese/etvbharatassamese-epaper-dh6b381d65c3344bbcad9a06ee28b4ab2a/boma+nikshepeve+dhbans+kva+hl+santvasabadiv+aatmagopanasthali-newsid-n400000150"
|
67 |
+
}
|
68 |
+
```
|
69 |
+
### Data Fields
|
70 |
+
|
71 |
+
- id: unique identifier for the artilce on DailyHunt. This id will be used to recreate the dataset.
|
72 |
+
- langCode: ISO 639-1 language code
|
73 |
+
- source_url: the url that points to the article on the website of the original publisher
|
74 |
+
- dh_url: the url that points to the article on DailyHunt
|
75 |
+
|
76 |
+
### Data Splits
|
77 |
+
|
78 |
+
From every language, we randomly sample 10,000 articles each for validation and testing. We also ensure that at least 80% of a language’s data is available for training.
|
79 |
+
Therefore, if a language has less than 100,000 articles, we restrict its validation and test splits to 10% of its size.
|
80 |
+
|
81 |
+
We also create a SMALL training set by limiting the number of articles from each language to 100K.
|
82 |
+
This SMALL training set with a size of 1.3M is used in all our fine-tuning experiments.
|
83 |
+
|
84 |
+
### Data Recreation
|
85 |
+
To recreate the dataset, follow this [README file](https://github.com/rahular/varta/tree/main/crawler#README.md).
|
86 |
+
|
87 |
+
### Source Data
|
88 |
+
https://m.dailyhunt.in/
|
89 |
+
|
90 |
+
|
91 |
+
### Limitations
|
92 |
+
|
93 |
+
This work is mainly dedicated to the curation of a new multilingual dataset for Indic languages, many of which are low-resource languages. During data collection, we face several limitations that can potentially result in ethical concerns.
|
94 |
+
Some of the important ones are mentioned below:
|
95 |
+
- Our dataset contains only those articles written by DailyHunt’s partner publishers. This has the potential to result in a bias towards a particular narrative or ideology that can affect the representativeness and diversity of the dataset.
|
96 |
+
- Another limitation is the languages represented in V ̄arta. Out of 22 languages with official status in India, our dataset has only 13. There are 122 major languages spoken by at least 10,000 people and 159 other languages which are extremely low-resourced. None of these languages are represented in our dataset.
|
97 |
+
- We do not perform any kind of debiasing on Varta. This means that societal and cultural biases may exist in the dataset, which can adversely affect the fairness and inclusivity of the models trained on it.
|
98 |
+
|
99 |
+
|
100 |
+
### Licensing Information
|
101 |
+
|
102 |
+
Apache License 2.0
|
103 |
+
|
104 |
+
### Citation Information
|
105 |
+
|
106 |
+
```
|
107 |
+
@misc{aralikatte2023varta,
|
108 |
+
title={V\=arta: A Large-Scale Headline-Generation Dataset for Indic Languages},
|
109 |
+
author={Rahul Aralikatte and Ziling Cheng and Sumanth Doddapaneni and Jackie Chi Kit Cheung},
|
110 |
+
year={2023},
|
111 |
+
eprint={2305.05858},
|
112 |
+
archivePrefix={arXiv},
|
113 |
+
primaryClass={cs.CL}
|
114 |
+
}
|
115 |
+
```
|