Masaru Isonuma is a JSPS overseas research fellow (postdoc) at the University of Edinburgh, working with Ivan Titov. He is currently interested in training data of language models (LMs): which training data affects LMs' performance, what LMs essentially learn from training data, and how to optimize training data. Before joining the Edinburgh NLP group, he obtained his PhD from the University of Tokyo, studying unsupervised summarization and topic models under the supervision of Ichiro Sakata, Junichiro Mori, and Danushka Bollegala.
Work Experiences
Apr. 2023-Present: JSPS Overseas Research Fellow, The University of Edinburgh
Apr. 2024-Present: Visiting Assistant Professor, National Institute of Informatics
Apr. 2023-Present: Visiting Researcher, The University of Tokyo
Apr. 2022-Mar. 2023: Project Assistant Professor, The University of Tokyo
Nov. 2021-Mar. 2022: Postdoctoral Researcher, The University of Tokyo
Apr. 2020-Oct. 2021: JSPS Research Fellow, The University of Tokyo
Apr. 2017-Sep. 2018: Professional Staff, Industrial Growth Platform, Inc.
Education & Qualifications
Oct. 2018-Sep. 2021: Ph.D. (Engineering), The University of Tokyo
Apr. 2015-Mar. 2017: M.S., The University of Tokyo
Apr. 2011-Mar. 2015: B.S., The University of Tokyo
Selected Publications
Masaru Isonuma, Ivan Titov. Unlearning Traces the Influential Training Data of Language Models. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (ACL, long paper), pp.6312-6325, 2024. [paper, code]
Tetsu Kasanishi, Masaru Isonuma, Junichiro Mori, Ichiro Sakata. SciReviewGen: A Large-scale Dataset for Automatic Literature Review Generation. Findings of the 61st Annual Meeting of the Association for Computational Linguistics (Findings of ACL, long paper), pp.6695–6715, 2023. [paper, code]
Nozomu Miyamoto, Masaru Isonuma, Sho Takase, Junichiro Mori, Ichiro Sakata. Dynamic Structured Neural Topic Model with Self-Attention Mechanism. Findings of the 61st Annual Meeting of the Association for Computational Linguistics (Findings of ACL, long paper), pp.5916–5930, 2023. [paper, code]
Masaru Isonuma, Junichiro Mori, Ichiro Sakata. Differentiable Instruction Optimization for Cross-Task Generalization. Findings of the 61st Annual Meeting of the Association for Computational Linguistics (Findings of ACL, Long Paper), pp.10502–10517, 2023. [paper, code]
Naomi Sato, Masaru Isonuma, Kimitaka Asatani, Shoya Ishizuka, Aori Shimizu and Ichiro Sakata. Lexical Entailment with Hierarchy Representations by Deep Metric Learning. Findings of the 2022 Conference on Empirical Methods in Natural Language Processing (Findings of EMNLP, short paper), pp.3517–3522, 2022. [paper]
Masaru Isonuma, Junichiro Mori, Danushka Bollegala, Ichiro Sakata. Unsupervised Abstractive Opinion Summarization by Generating Sentences with Tree-Structured Topic Guidance. Transactions of the Association for Computational Linguistics (TACL), vol.9, pp.945–961, 2021. [paper, code]
Masaru Isonuma, Junichiro Mori, Danushka Bollegala, Ichiro Sakata. Tree-Structured Neural Topic Model. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL, Short Paper), pp.800-806, 2020. [paper, code]
Masaru Isonuma, Junichiro Mori, Ichiro Sakata. Unsupervised Neural Single-Document Summarization of Reviews via Learning Latent Discourse Structure and its Ranking. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL, Long Paper), pp.2142–2152, 2019. [paper, code]
Masaru Isonuma, Toru Fujino, Junichiro Mori, Yutaka Matsuo, Ichiro Sakata. Extractive Summarization using Multi-Task Learning with Document Classification. Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing (EMNLP, Long Paper), pp.2101–2110, 2017. [paper]
Awards
Best Paper Award, The 30th Annual Meeting of the Association for Natural Language Processing, 2024. (first author, 1/599 papers)
JSAI Annual Conference Award, The 37th Annual Conference of the Japanese Society for Artificial Intelligence, 2023. (first author, 38/954 papers)
Outstanding Paper Award, The 29th Annual Meeting of the Association for Natural Language Processing, 2023. (co-author, 13/579 papers)
Yamashita SIG Research Award, Information Processing Society of Japan, 2022. (first author, 1/33 papers)
Dean's Award for the Best Doctoral Thesis, Graduate School of Engineering, The University of Tokyo, 2022.
Young Researcher Encouragement Award, The 27th Annual Meeting of the Association for Natural Language Processing, 2021. (first author, 11/247 papers)
Excellence Award, Special Interest Group of Natural Language Processing, 2020. (first author, 2/22 papers)
Grants & Fellowships
JSPS Overseas Research Fellowships
JSPS Grant-in-Aid for Early-Career Scientists
JST ACT-X Acceleration Phase (Frontier of Mathematics and Information Science)
JST ACT-X (Frontier of Mathematics and Information Science)
JSPS Research Fellowships for Young Scientists (Doctoral Course 2)
TOYOTA/Dwango Scholarship for Advanced Artificial Intelligence Researcher
JSPS Overseas Challenge Program for Young Researchers (declined due to COVID-19)
UTokyo-TOYOTA Study Abroad Scholarships for Advanced Artificial Intelligence Researcher (declined due to COVID-19)
SEUT Doctoral Student Special Incentives Program (Type A)