Bert-based Braille Summarization of Long Documents

Availability of accurate document summaries is of paramount importance in our lives, though there is a dearth of such options for the visually challenged people. With a large volume of existing text summarization work dealing with short documents, this work aims to tackle the problem of large document summarization in braille. We propose a novel domain and genre agnostic approach to document summarization in braille using a BERT extractive summarizer. Experimental evaluation of our approach involved a comparison of the generated and reference summaries using ROUGE metrics. Given the lack of robust convertors from English to Braille, another key contribution of our work is the development of a novel English & Braille interconversion library to facilitate further research in this domain.