Natural Language Processing based Cross Lingual Summarization

Abstract

Natural Language Processing (NLP) is a domain that programs machines to interpret/ comprehend human language like human beings do. Application of NLP to summarization and machine translation has been accelerating from past years. One such application, Cross-lingual summarization (CLS), produces a summary in a particular language (target language) for an input document in a different language (source language). CLS can help people understand an article's gist written in an unfamiliar language by translating the summary to their familiar language. Current methods divide CLS into two methods: Summarization (Abstractive or Extractive) and Translation. Although there has been a tremendous amount of progress in abstractive summarization in recent years, most research focuses on monolingual summarization because of lack of high-quality multilingual resources. While there have been a few studies to address the lack of resources for cross-lingual summarization, the datasets employed are very limited in size. This paper presents the Cross-Lingual Summarization from English to Kannada by summarizing English ne articles and translating the summaries to Kannada. Traditional CLS methods are pipeline-based i.e., it either involves summarization and then translation or vice versa. The summarization is followed by translation, hence it is termed as 'Late Translation'.