H-QFS : hybrid query-focused for multi-perspective scientific document summarization (MUP) | |
| Author | Nopphawan Nurnuansuwan |
| Call Number | AIT Thesis no.DSAI-23-10 |
| Subject(s) | Natural language processing (Computer science) Information retrieval |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Data Science and Artificial Intelligence |
| Publisher | Asian Institute of Technology |
| Abstract | Scientific document summarization often relies on datasets featuring single gold summaries per document, potentially overlooking the diverse perspectives inherent in real world scientific content. Despite efforts made by the Multi-Perspective Scientific Document Summarization (MuP) dataset to encompass diverse viewpoints through multi perspective summaries, a persistent lack of clarity regarding these perspectives remains evident. Existing approaches have not adequately prioritized the intricate task of multi perspective summarization. In response to this challenge, our paper introduces the Hybrid Query-Focused Summarization framework (H-QFS), a novel approach explicitly crafted to address the dual challenges of general summarization and Query-Focused Summarization (QFS) in scientific documents. Our H-QFS framework not only excels in the QFS task on a synthetic query-focused validation dataset, achieving state-of-the-art performance with the highest Rouge-1, Rouge-2, Rouge-L, and Rouge-average scores, but also competes favorably with existing methods and baselines in general summa rization on a blind test set, securing the 2nd place in Rouge-1 in the blind test set. This underscores the potential of query-focused summarization in extracting diverse perspectives from scientific documents. Our study makes three main contributions: Modification of Query-Focused Summarization (QFS) to effectively summarize multi perspective scientific papers, proposing a versatile Hybrid Query-Focused Summariza tion (H-QFS) framework that works for both general summaries and specific queries, and developing a General Summarization (GS) framework and H-QFS that perform well in blind tests. Our experiments and frameworks are available on https://github.com/ Thetang-145/MuP_Query_Focused_Summarization |
| Year | 2023 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Data Science and Artificial Intelligence (DSAI) |
| Chairperson(s) | Chaklam Silpasuwancha; |
| Examination Committee(s) | Dailey, Matthew N.; |
| Scholarship Donor(s) | AIT Scholarships; |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2023 |