JFinTEB: Japanese Financial Text Embedding Benchmark
A benchmark for evaluating text embeddings in Japanese finance.
RESEARCHER & QUANT ANALYST
Masahiro Suzuki
鈴木 雅弘 / Ph.D.
Research interests
Natural language processing for finance.
My research spans domain-specific language models, financial text analysis, and open datasets and benchmarks.
Quant AnalystAmova Asset Management
Project Assistant ProfessorKeio University, SDM
News
Appointed Project Assistant Professor at Keio University, SDM.
Our paper on subjective financial knowledge was accepted at IIAI AAI 2026 (full paper).
Provided research assistance for the Netflix series Romantics Anonymous (episodes 1, 7, and 8).
Received my Ph.D. in Engineering from the University of Tokyo.
Selected papers
All publications on Google Scholar ↗Economy Watchers Survey Provides Datasets and Tasks for Japanese Financial Domain
Datasets and tasks built from Japan’s Economy Watchers Survey.
Constructing and analyzing domain-specific language model for financial text mining
Domain-specific language modeling for financial text analysis.
Open resources
GitHub ↗BENCHMARK
JFinTEB ↗
Text embedding evaluation for Japanese finance.
DATASET
Economy Watchers Survey ↗
Japanese financial and economic text.
DATASET
llm-japanese-dataset ↗
Japanese chat data for instruction tuning of language models.
Models & toolsJapanese DeBERTaV2 ↗jptranstokenizer ↗
Selected academic service
- 2026–NLC, IEICEAssistant Secretary
- 2025–2026IEEE CIFEr 2026Organizer
- 2026SIGIR, Resources Papers TrackProgram Committee
Selected honors
- 2024Research AwardNLC, IEICE
- 2023Honorable Mention & ELYZA AwardYANS Symposium
- 2019Best Paper AwardCDEC, ICDM Workshops
Ph.D. in Engineering, the University of Tokyo, 2025. CV, including scholarships and appointments ↗ · researchmap ↗