---
license: mit
task_categories:
- text-classification
- zero-shot-classification
- token-classification
language:
- te
- en
tags:
- telugu
- language
- nlp
pretty_name: Telugu Chandassu
size_categories:
- 1K<n<10K
---
![plot](assets/banner.png)

First ever dataset for "Chandassu", metrical poetry in Telugu Language (తెలుగు చంధస్సు). </br>
arXiv Preprint: [Read Here](https://arxiv.org/abs/2510.01233)

Our algorithm achieves **91.73% accuracy** on the proposed Chandassu Score.

## Getting Started

```py
from datasets import load_dataset
data= load_dataset( "BodduSriPavan111/chandassu", split= "train" )
df= data.to_pandas()
print( df.shape )
```

## Description

Chandassu dataset comprises **4,651** carefully curated Telugu padyams spanning three primary prosodic classes (Vruttamu, Jaathi, Vupajaathi) and eight distinct types (Vutpalamaala, Champakamaala, Saardulamu, Mattebhamu, Kandamu, Teytageethi, Aataveladi, Seesamu).
Each entry in the dataset contains the following structured attributes:

- **type:** Type of Padyam </br>
- **padyam:** Padyam text </br>
- **class:** Class of padyam </br>
- **satakam:** Satakam that the padyam belongs to </br>
- **lg_data:** LaghuvuGuruvu data </br>
- **chandassu_score:** Proposed metric to evaluate padyam with given type</br>
- **n_aksharaalu_score:** Ratio of number of aksharam tokens present in given input text to the number of expected aksharam tokens in particular padyam. </br>
- **n_paadalu_score:** Ratio of number of paadams found in the given input text to the expected number of paadams in the padyam. </br>
- **gana_kramam_score:** Ratio of total number of ganams matched sequentially to the total expected number of ganams for the padyam. </br>
- **yati_score:** : Ratio of number of paadams of yati match to the total number of paadams expected to satisfy yati in the padyam. </br>
- **prasa_score** Ratio of frequency of modal prasa aksharam token to the expected number of paadams for the padyam. </br>


## Eager to Contribute?
Any data, features, refactoring or **your innovative thought**, please check <a href= "https://github.com/BodduSriPavan-111/chandassu/blob/main/CONTRIBUTING.md">here</a>.

## Acknowledgements
Special thanks to Sesha Sai Vadapalli and Kalepu Nagabhushana Rao, maintainers of [andhrabharati.com](https://andhrabharati.com/). </br>
Appana Mohan Naga Phani Kumar for proofreading our article. </br>
Sincere gratitude to our parents and family members for their continuous support throughout this work. </br>
This work was undertaken with the grace of Sri Ramalinga Chowdeswari Devi.

## Citation
>@misc{pavan2025computationalsociallinguisticstelugu, </br>
>      title={Computational Social Linguistics for Telugu Cultural Preservation: Novel Algorithms for Chandassu Metrical Pattern Recognition}, </br>
>      author={Boddu Sri Pavan and Boddu Swathi Sree}, </br>
>      year={2025}, </br>
>      eprint={2510.01233}, </br>
>      archivePrefix={arXiv}, </br>
>      primaryClass={cs.CL}, </br>
>      url={https://arxiv.org/abs/2510.01233},  </br>
>}

## Thank You !