HINDI TEXT SUMMARIZER USING ABSTRACTIVE AND EXTRACTIVE TECHNIQUE

Abstract

Text summarization is a significant process in retrieving the information, supporting the generation of concise summaries from original text documents. This paper looks at the importance of text summarization in many different uses those being Article Proofreading, Online Visibility Enhancement, Corporate Assessment, and Competitive Analysis. Both the two main text summarization strategies, a creative and a concise one, will be debated. Summary based on an extractive method depends on identifying important sentences and words from the text, based on high scores. On the other hand, an abstractive summary can contain some new expressions which are not presented in the unduplicated text. The paper is based on an extractive summarization model that has been developed using TFIDF. TFIDF approaches, Clustering-based techniques, and ML approaches are reported to be efficient for extractive text summarization and abstract based model is developed using BERT technique its ability to perceive language structures, sentence relationships, and to understand meaning gets significantly improved in a similar patterns of humans. Having been pre-trained, BERT will be able to do fine-tuning for numerous downstream NLP tasks, e.g., summarizing the text, NER, and Q and A. One of its strengths is being able to produce contextualized vector representations of words, where the semantical association of a word is dynamically modulated by the context. The study reviews the current approaches and models for Hindi text summarization, concentrates on the semantic relatedness principle and the keyword feature extraction.