Publications
WORKSHOP (INTERNATIONAL) University of Tsukuba Team at the TREC 2023 Interactive Knowledge Assistance Track
LINGZHEN ZHENG (Tsukuba univ.), KAIYU YANG (Tsukuba univ.), HAITAO YU (Tsukuba univ.), SUMIO FUJITA, HIDEO JOHO (Tsukuba univ.)
The Thirty-Second Text REtrieval Conference (TREC 2023)
March 01, 2024
In this paper, we present our approach employed in the four automatic submission runs for the TREC 2023 iKAT (Interactive Knowledge Assistance Track). This track comprises three subtasks: passage ranking, response generation, and PTKB statement ranking (Personal Text Knowledge Base). Our comprehensive multi-stage pipeline for this task encompasses query rewriting, PTKB statement ranking, passage retrieval, passage reranking, and response generation. In our four submissions, we employed fine-tuned pre-trained T5-CANARD for query rewriting, a combination of BERT, RankGPT, and MonoT5 for PTKB statement ranking, and LLMs, RankGPT, and MonoT5 for passage reranking.
Paper :
University of Tsukuba Team at the TREC 2023 Interactive Knowledge Assistance Track
(external link)