Query-Based Summarization using Reinforcement Learning and Transformer Model

Query-based summarization is an interesting problem in the text summarization field. The reinforcement learning technique is popular for robotics and become accessible for the text summarization problem in the recent years. The lack of significant works with reinforcement learning in this field inspired us to solve the query-based summarization problem using this technique. While doing so, we have introduced a different approach for sentence ranking and clustering to avoid the existence of redundant data in the summaries. We have proposed an unsupervised extractive summarization method, which provides some state-of-the-art results. We also developed two abstractive multi-document summarization models using reinforcement learning technique and the transformer model. We consider the importance of information coverage and diversity under a fixed sentence limit for our summarization models. We have done several experiments for our proposed models, which bring significant results across different evaluation metrics.