Koki Shibata (Kato Laboratory)

Since entering the master’s program in Informatics, I have been actively engaged in research activities at the Kato Laboratory (https://kasys.slis.tsukuba.ac.jp/) and presented my research at both domestic and international conferences. First, I presented the research I had been working on since my undergraduate days at the ICADL 2022, a peer-reviewed international conference on digital library, under the title of “Active Learning for Efficient Partial Improvement of Learning to Rank.” This study proposes a way to fix problematic search results without changing other search results in the search algorithm.

In addition, I participated in an internship at Hitachi, Ltd. and presented a research work entitled “How do different Japanese tokenizers perform on downstream tasks?” (original title: 日本語Tokenizer の違いは下流タスク性能に影響を与えるか？) at the domestic conference of NLP 2023. This study investigates the impact of tokenizers on the performance of large-scale language models in Japanese. Furthermore, after brushing up on the presented work, we plan to present it at the ACL SRW 2023, a peer-reviewed international workshop on natural language processing, under the title of “How do different tokenizers perform on downstream tasks in scriptio continua languages?: A case study in Japanese.”

This degree program offers a good support and environment for students’ research activities. Let’s take advantage of this support and environment to learn and grow together. I sincerely welcome your application and becoming a member of our degree program.

Koki Shibata, Kato Laboratory,
Second-year master student at Master’s Program in Informatics

Note: Student affiliation and grade are current as of June 2023.