Hanyoung and Yanggee with their poster in KSC 2023 (Busan, South Korea)Abstract
In information retrieval, traditional models like TF-IDF and BM25 have been used for search engine queries. However, with the rise of deep learning in natural language processing, DPR-BERT has significantly improved accuracy in this field. However, DPR-BERT has limitations related to precise retrieval concerning perspectives. In this paper, to address these limitations, we propose a new architecture that extracts perspective and incorporates them into the retrieval process. In this process, experiments were conducted by modifying the method of obtaining similarity and the training objective function defined by the original DPR to align with our goals. Our experimental results confirm that incorporating perspective during retrieval significantly improves the process compared to not doing so.
Hanyoung Kim
Ph.D. Student
I’m the last M.S. in Financial Engineering graduate managed from Econ department!
