README.md

---
license: mit
task_categories:
- text2text-generation
language:
- bn
pretty_name: BanglaTLit
size_categories:
- 10K<n<100K
---

# BanglaTLit: A Benchmark Dataset for Back-Transliteration of Romanized Bangla

[![arXiv](https://img.shields.io/badge/Code-farhanishmam/BanTH-blue?logo=GitHub)](https://github.com/farhanishmam/BanglaTLit)

## Dataset Overview

- **BanglaTLit-PT:** A pre-training corpus with 245727 transliterated or romanized Bangla samples for further pre-training language models.
- **BanglaTLit:** Subset of the BanglaTLit-PT dataset containing 42705 romanized Bangla and its corresponding Bangla back-transliteration pairs.

## Data Description

| **Column Title**     | **Description**                                                                 |
|----------------------|----------------------------------------------------------------------------------|
| `id`                 | A unique identifier for each data point in the dataset.                         |
| `text_transliterated`| The transliterated (romanized) Bengali text input that needs to be back-transliterated. |
| `text_bengali`       | The corresponding Bengali text that is the correct back-transliteration output.  |

## Summary Statistics

 | Statistic              | Transliterated    | Back-Transliterated   |
 |-------------------------|-------|-------|
 | Mean Character Length    | 59.24 | 58.28 |
 | Max Character Length     | 1406  | 1347  |
 | Min Character Length     | 3     | 4     |
 | Mean Word Count          | 10.35 | 10.51 |
 | Max Word Count           | 212   | 226   |
 | Min Word Count           | 2     | 2     |
 | Unique Word Count        | 81848 | 60644 |
 | Unique Sentence Count    | 42705 | 42471 |