sohomghosh/IndicFinNLP_FinancialNatural_Language_Processing_for_Indian_Languages

---
license: cc-by-nc-sa-4.0
task_categories:
- text-classification
language:
- bn
- hi
- te
tags:
- finance
size_categories:
- 10K<n<100K
---
# IndicFinNLP

This repository contains dataset mentioned in the paper, "IndicFinNLP: Financial Natural Language Processing for Indian Languages" @ LREC-COLING 2024

Tasks:

Exaggerated Numeral Detection
Sustainability Assessment,
ESG Theme Determination
Languages:
Hindi, Bengali, Telugu

Source:

Budget speeches of Hindi, Bengali, and Telugu speaking states (Punjab, Uttarakhand, Haryana, West Bengal, Telangana, and Andhra Pradesh) starting from the year 2011 till 2023.
Financial texts filtered from the Samanantar corpus (Rameshet al., TACL 2022).
Translated the existing dataset proposed by Kangand El Maarouf (FinSim4-ESG FinNLP-2022) from English to Indian languages (Hindi, Bengali, and Telugu). Retained only the high quality ones.
Translated the existing dataset proposed by Chen et al. 2023 (FinNLP-2023 ML-ESG) from English to Indian languages (Hindi, Bengali, and Telugu). Manually verified and corrected the translations
wherever needed.

## Resources

**Task-1 Metadata** <br>
------------------- <br>
| column name | explanation | example |
|----------------|--------------------------------------------------------------------|-------------------------------|
| indic | financial text in indic language | যা টাকায় ১০ কোটি টাকারও বেশি। |
| number_english | number present in indic text after translating it to English | 10 |
| number_indic | number present in indic text | ১০ |
| start_posn | starting position of the number in indic text | 9 |
| end_posn | ending position of the number in indic text | 11 |
| language | indic language in which the text is present (hindi/bengali/telugu) | bengali |
| magnitude | magnitude of the number | 1 |

**Task-2 Metadata** <br>
------------------- <br>
| column name | explanation | example |
|----------------|----------------------------------|-----------------------------------------------------------------|
| sentence_indic | financial text in indic language | 2019 में, हवाई यात्रा हमारे अपने कार्बन फुटप्रिंट का लगभग 38 प्रतिशत थी। |
| label | sustainable or unsustainable | unsustainable |
| language | indic language | hindi |

**Task-3 Metadata** <br>
------------------- <br>
| column name | explanation | example |
|------------------|------------------------------|-----------------------------------------------------------------------------------------|
| URL | url of the news title | https://www.esgtoday.com/abn-amro-to-align-lending-investment-portfolios-with-net-zero/ |
| news_title_indic | news title in indic language | రుణాలను సమలేఖనం చేయడానికి, నికర సున్నాతో పెట్టుబడి దస్త్రాలను సమలేఖనం చేయడానికి ABN AMRO |
| ESG_Theme | ESG theme of the news title | climate change |
| language | indic language | telugu |

```bibtex
@INPROCEEDINGS{ghosh2024,
author={Ghosh, Sohom and Majhi, Arnab and Narayana, Aswartha and Naskar, Sudip Kumar},
booktitle={Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)},
title={IndicFinNLP: Financial Natural Language Processing for Indian Languages},
month={05},
year={2024},
pages={9010–9018},
url={https://aclanthology.org/2024.lrec-main.789/}
}
```

---
license: cc-by-nc-sa-4.0
---

原始信息汇总

数据集概述

数据集名称

IndicFinNLP

数据集来源

• 预算演讲文本来自印度语地区（旁遮普邦、北阿坎德邦、哈里亚纳邦、西孟加拉邦、特伦甘纳邦和安得拉邦），时间跨度为2011年至2023年。
• 从Samanantar语料库中筛选的金融文本。
• 翻译自Kang和El Maarouf提出的FinSim4-ESG FinNLP-2022数据集，以及Chen et al. 2023提出的FinNLP-2023 ML-ESG数据集，从英语翻译至印度语（印地语、孟加拉语、泰卢固语）。

数据集任务

1. 夸大数字检测
2. 可持续性评估
3. ESG主题确定

数据集语言

• 印地语
• 孟加拉语
• 泰卢固语

数据集规模

10K<n<100K

数据集许可证

cc-by-nc-sa-4.0

数据集元数据

Task-1 Metadata

列名	解释	示例
indic	印度语金融文本	যা টাকায় ১০ কোটি টাকারও বেশি।
number_english	印度语文本中的数字翻译成英语	10
number_indic	印度语文本中的数字	১০
start_posn	数字在印度语文本中的起始位置	9
end_posn	数字在印度语文本中的结束位置	11
language	文本使用的印度语（印地语/孟加拉语/泰卢固语）	bengali
magnitude	数字的量级	1

Task-2 Metadata

列名	解释	示例
sentence_indic	印度语金融文本	2019 में, हवाई यात्रा हमारे अपने कार्बन फुटप्रिंट का लगभग 38 प्रतिशत थी।
label	可持续或不可持续	unsustainable
language	文本使用的印度语	hindi

Task-3 Metadata

列名	解释	示例
URL	新闻标题的网址	https://www.esgtoday.com/abn-amro-to-align-lending-investment-portfolios-with-net-zero/
news_title_indic	印度语新闻标题	రుణాలను సమలేఖనం చేయడానికి, నికర సున్నాతో పెట్టుబడి దస్త్రాలను సమలేఖనం చేయడానికి ABN AMRO
ESG_Theme	新闻标题的ESG主题	climate change
language	文本使用的印度语	telugu

数据集相关文献

• Ghosh, Sohom and Majhi, Arnab and Narayana, Aswartha and Naskar, Sudip Kumar. "IndicFinNLP: Financial Natural Language Processing for Indian Languages". Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), May 2024, pp. 9010–9018.