Skip to content
All work

YANS2026 Natural Language Processing Symposium: Building a Japanese Design Gazette Dataset and Cross-Modal Benchmarks

Research that organises images, bibliographic information, and documents from Japanese design gazettes and defines evaluation tasks for cross-modal design understanding. Presented at the 21st Natural Language Processing Symposium for Young Researchers (YANS2026).

Natural Language Processing Symposium for Young Researchers (YANS2026)
Reo Hakuta presenting design gazette dataset research at the YANS2026 natural language processing symposium

This research transforms bibliographic information, descriptions, classifications, and drawings from multiple viewpoints in Japanese design gazettes into a form that can be used across machine-learning tasks. I presented it as a poster at the 21st Natural Language Processing Symposium for Young Researchers (YANS2026) under the Japanese title “クロスモーダル意匠理解に向けた日本語意匠公報データセット構築とベンチマーク設計.”

Presentation Materials

The poster on the Japanese design-gazette dataset presented at YANS2026
Image generated from the research poster PDF published on Google Drive.

The presentation materials are available in the public Google Drive folder. The research poster PDF can also be viewed directly.

YANS2026 Hackathon ICL Division Excellence Award

I joined team ICL-c, 「牛タン正直ナメてた」, at the cross-disciplinary hackathon on 16 August. The team received an Excellence Award in the ICL division.

Team photograph taken after receiving the Excellence Award in the ICL division of the YANS2026 hackathon
Team ICL-c after receiving the Excellence Award in the ICL division.

The SB Intuitions TECH BLOG report on YANS2026 featured the presentation for which I was first author as one of the event's notable presentations.

Next work

Locker.ai: Lost Property Management with an LLM and Smart Lockers