Evaluation of Extractive and Abstract Methods in Text Summarization

Text summarization has become very essential tool to record important points and has been used by several websites and applications to lessen length, difficulty, and to preserve the vital information of the original file. The requirement on well-organized and useful text summarization of the website content, news feed and other kinds of legal documents with judgments and predilection is the demand of the present requirement. Hence several attempts have been made to automate the summarizing process. The recent development and state of the art models in natural language processing demonstrated outstanding results in text summarization, however major focus of these analysis was on large dataset with large parameters. This study’s primary purpose is to evaluate the performance of ensemble abstractive and extractive models on text summarization. Combined core of BERT and PEGASUS models’ output were applied to LexRank model on News Summary dataset to evaluate the performance through ROUGE metric. The results showed the performance of combined and ensemble model is better than individual performance.