Research Themes
Research themes
Humans communicate and share knowledge through speech, writing, figures, tables, spaces, media,
and artificial intelligence. My research aims to understand the structures behind communication
and knowledge transfer, and to develop practical intelligent information processing technologies.
1. LLM analysis and knowledge-/constraint-aware AI systems
I analyze large language model behavior using information-theoretic approaches, including
information spectrum theory, and study how side information, external knowledge, and constraints
affect generation behavior.
2. Recommendation, RAG, and knowledge-graph-based information access
I study knowledge-augmented recommendation with lightweight LLMs, retrieval-augmented generation,
and information access methods that incorporate user knowledge, venue knowledge, and structured knowledge.
3. Multimodal document understanding and structured knowledge extraction
I work on understanding visually rich documents that include text, tables, figures, and layouts,
including text-table relatedness, joint triple extraction, and table cell entity annotation.
4. Wireless propagation and physics-constrained generative modeling for the 6G era
I study path loss map estimation, radio propagation modeling, wave simulation verification,
and physics-constrained generative modeling using vision transformers and diffusion models.
5. Earlier work on acoustics, speech, dialogue, and quantitative linguistics
My earlier research includes room acoustic analysis, robust speech recognition under noise and reverberation,
speech enhancement, spoken dialogue systems, and quantitative analysis of classical Japanese literature.