HinExSum: Hindi extractive text summarization using word embeddingsImage 999

HinExSum is an open-source Python library for extractive Hindi summarization. To the best of our knowledge, it delivers the first openly released reference implementation of Mohd et al.’s (2020) word-embedding-based summarization method, re-engineered end-to-end for Hindi. The package incorporates danda-aware sentence segmentation, light stemming, Hindi stop-word removal, semantic big-vector generation using fastText Hindi embeddings, k-means sentence clustering, and seven-feature sentence ranking with redundancy reduction. HinExSum offers independently reusable components for low-resource Indic NLP: a Devanagari-safe ROUGE-1/2/L/SU4 scorer and curated Hindi lexical resources. An evaluation harness with paired bootstrap confidence intervals supports benchmarking on the XL-Sum and ILSUM Hindi datasets.