Cross lingual summarization improves Marathi to English text summaries
Cross Lingual Abstractive Summarization for Marathi
Computation and LanguageArtificial Intelligence
Summary
Summarizing text in different languages is hard, especially for Indian languages like Marathi. The authors worked on making summaries that translate Marathi text into English summaries using advanced machine learning models. By combining existing Marathi data and a smaller translated dataset, they improved the quality of these summaries better than previous methods. Their work mainly focused on translating Marathi to English summaries but discussed the reverse direction as well.
What this means in practice
- •For multilingual content platforms: Generate English summaries from Marathi articles to improve content accessibility and reach for wider audiences.$Commercial implications: Enables content platforms to offer better cross-lingual summarization features, attracting more users and advertisers.
- •For local news aggregators: Provide concise English summaries of Marathi news for readers who do not understand Marathi.
Authors
Asang Dani, Shailesh R Sathe
Abstract
Text summarization is an important task for advancing natural language processing (NLP) research and applications in Indic languages. Despite progress in extractive and abstractive summarization, both monolingual summarization (MS) and cross-lingual summarization (CLS) remain relatively underexplored for these languages. Recent advances in Transformer architectures, pretrained language models (PLMs), and dataset availability offer new opportunities to address this gap. In this work, we construct cross-lingual training data from the monolingual Marathi MahaSum corpus and combine it with the smaller cross-lingual PMIndiaSum dataset to fine-tune IndicTrans2, an Indic neural machine translation model. We evaluate the resulting models using ROUGE, chrF and chrF++ metrics, focusing primarily on Marathi-to-English summarization due to computational constraints. We also discuss the applicability of our approach to the English-to-Marathi direction. Our experiments demonstrate that leveraging a pretrained translation model yields substantially better summarization performance than the IndicBART and mBART-50 based baselines reported in previous work on cross-lingual summarization.