L3Cube-MahaPOS: A Marathi Part-of-Speech Tagging Dataset and BERT Models

2026-06-23

AI Summary

This study addresses the critical scarcity of high-quality part-of-speech (POS) annotated data and standardized evaluation benchmarks for Marathi, which has significantly hindered natural language processing (NLP) research for the language. To bridge this gap, the authors construct the first large-scale, manually curated dataset comprising 32,354 sentences of news text, annotated with 16 universal dependency POS tags. Rigorous preprocessing—including Unicode normalization, Devanagari-aware tokenization, and noise filtering—ensures data consistency. Comprehensive evaluations using HMM, CRF, BiLSTM, CharCNN, and transformer-based models such as MuRIL and MahaBERT-v2 demonstrate that the best-performing model achieves a token-level accuracy of 88.67% and a macro-averaged F1 score of 81.67% across 15 tags, substantially advancing foundational resources for Marathi NLP.

Abstract

Part-of-Speech (POS) tagging is a foundational NLP task underpinning machine translation, information extraction, and syntactic parsing. Despite Marathi being spoken by over 83 million people and ranking among the top twenty most spoken languages worldwide, it remains severely under-resourced in annotated corpora and standardised evaluation benchmarks. Marathi presents unique challenges for computational modelling owing to its rich morphology, relatively free word order, lack of capitalisation conventions, and pervasive code-mixing with Hindi and English. We introduce L3Cube-MahaPOS, a gold-standard POS tagging dataset for Marathi comprising 32,354 manually annotated sentences drawn from news text. Annotation was performed entirely manually by a team of Marathi-proficient annotators following a 16-tag Universal Dependencies-aligned scheme. A structured preprocessing pipeline covering Unicode normalisation, Devanagari-aware tokenisation, and noise filtering ensures label consistency across all splits. We benchmark the dataset across six model families spanning HMM, CRF, BiLSTM, BiLSTM+CharCNN, MuRIL, and the Marathi-specific transformer MahaBERT-v2. The best system achieves 88.67\% token-level accuracy and a macro-F1 of 81.67% over 15 evaluated tag classes. We release the dataset, annotation guidelines, and trained model checkpoints to foster further research in Marathi NLP.

Problem

• Marathi
• Part-of-Speech tagging
• low-resource language
• annotated corpus
• NLP benchmark

Innovation

• Marathi POS tagging
• low-resource NLP
• BERT-based models
• Universal Dependencies
• Devanagari-aware tokenization