Improved Text Summarization with RL-Enabled PEGASUS and Siamese Network Evaluation

Text summarization is a fundamental down-stream task in NLP, a field of AI that focuses on the interactions between computers and humans through natural language. With the exponential growth of digital information and advancements in various fields of science, the need for summarizing scientific papers has increased drastically. Classical statistical methods, deep learning models, and transformer-based architectures are employed to extract concise summaries from large scientific datasets. In this paper, a comparative analysis of these techniques is conducted using standard benchmark datasets and evaluation metrics, such as ROUGE scores, to assess their performance in terms of summary quality, coherence, and informativeness. An RL-enabled PEGASUS model is proposed for slightly enhanced performance in scientific text summarization. Addressing the current lack of a proper evaluation metric for summarization tasks, a new Siamese Networks-based architecture is proposed to serve as a better evaluation metric for generative language models, including text-summarization models.