# L3Cube-MahaSTS: A Marathi Sentence Similarity Dataset and Models

## TL;DR

This paper introduces MahaSTS, a Marathi sentence similarity dataset, and MahaSBERT-STS-v2, a fine-tuned model that significantly improves similarity scoring in Marathi NLP tasks.

## Contribution

The paper provides the first large-scale Marathi sentence similarity dataset and a specialized BERT-based model for improved regression-based similarity scoring.

## Key findings

- MahaSTS dataset contains 16,860 annotated sentence pairs.
- MahaSBERT-STS-v2 outperforms other models like MahaBERT and IndicSBERT.
- Structured supervision enhances model stability in low-resource Marathi NLP.

## Abstract

We present MahaSTS, a human-annotated Sentence Textual Similarity (STS) dataset for Marathi, along with MahaSBERT-STS-v2, a fine-tuned Sentence-BERT model optimized for regression-based similarity scoring. The MahaSTS dataset consists of 16,860 Marathi sentence pairs labeled with continuous similarity scores in the range of 0-5. To ensure balanced supervision, the dataset is uniformly distributed across six score-based buckets spanning the full 0-5 range, thus reducing label bias and enhancing model stability. We fine-tune the MahaSBERT model on this dataset and benchmark its performance against other alternatives like MahaBERT, MuRIL, IndicBERT, and IndicSBERT. Our experiments demonstrate that MahaSTS enables effective training for sentence similarity tasks in Marathi, highlighting the impact of human-curated annotations, targeted fine-tuning, and structured supervision in low-resource settings. The dataset and model are publicly shared at https://github.com/l3cube-pune/MarathiNLP