AN EFFECTIVE METHOD COMBINING DEEP LEARNING MODELS AND REINFORCEMENT LEARNING TECHNOLOGY FOR EXTRACTIVE TEXT SUMMARIZATION

Abstract

Automatic text summarization is an important problem in natural language processing. Text summarization extracts the most important information from one or many source texts to generate a brief, concise summary that still retains main ideas, correct grammar and ensures the coherence of the text. With the application of machine learning techniques as well as deep learning models in automatic text summarization models gave summaries that were closely resemble human reference summaries. In this paper, we propose an effective extractive text summarization method by combining the deep learning models, the reinforcement learning technique and MMR method to generate the summary. Our proposed method is experimented on CNN dataset (English) and Baomoi dataset (Vietnamese) giving F1-score accuracy results with Rouge-1, Rouge-2, Rouge-L are 31.36%, 12.84%, 28.33% and 51.95%, 24.38%, 37.56%, respectively. The experimental results show that our proposed summarization method has achieved good results for English and Vietnamese text summarization.