‘Team KU_NLP (Professor Hak-su Kim)’, composed of Bo-eun Kim, Do-haeng Lee from the Department of Artificial Intelligence in the Graduate School of Konkuk University, and Ye-jin Lee in the senior year of Department of Applied Statistics, took the first place in the category of ‘Text Generation’ in ‘DialDoc 2021’, the international competition of natural language processing. (Thesis title: Document-Grounded Goal-Oriented Dialogue Systems on Pre-Trained Language Model with Diverse Input Representation) ▲ Bo-eun Kim ▲ Do-haeng Lee ▲ Ye-jin Lee
Hosted by the ACL (Association of Computer Linguistics) which is the top international academic organization in the field of natural language processing, ‘DialDoc 2021’ is an international competition participated by 20 leading universities and companies from home and abroad. The project is to create an AI chatbot that acquires knowledge on its own in large documents and generates appropriate responses after understanding the user’s natural language questions.The competition, which was held for about three months after the data release on this February, was divided into two parts; ‘Knowledge identification’ that finds appropriate knowledge from documents and ‘Text generation’ that generates responses from the knowledge. Team KU_NLP ranked the first in ‘Text generation’ and fifth in ‘Knowledge identification’.Do-haeng Lee said, “I found it interesting to research dialogue systems through document-based dialogue system that generates responses utilizing knowledge and information beyond simple dialogue systems,” adding that, “We plan to conduct in-depth research to develop the relationship between dialogue structure and knowledge documents with the ideas proposed in this competition.”Bo-eun Kim said, “It was a meaningful experience to apply various ideas while implementing document-based dialogue generation model,” and “I especially appreciate researchers who provided the statistics necessary for model generation and Prof. Hak-su Kim who guided us to get good results.”