Overview
In human-human and human-robot dialogue, participants do far more than exchange information: they adjust how they speak to their partner and the situation, acquire knowledge, and build social relationships. My research investigates the mechanisms underlying these dialogue behaviors through the lens of linguistic theories such as sociolinguistics and pragmatics, and develops dialogue intelligence that computationally supports them. I also focus on the intellectual property (IP) domain — which demands both sophisticated language use and deep expert knowledge — as a primary application field, developing natural language processing that supports professional practice. Beyond developing individual component technologies, I aim to realize an integrated framework in which a dialogue agent orchestrates these technologies, supporting access to expert knowledge and decision-making through dialogue with people.
User-Adaptive Dialogue Systems Grounded in Linguistic Theories
Humans unconsciously adjust their vocabulary, style, and manner of speaking depending on their dialogue partner and the situation. By formalizing sociolinguistic phenomena such as style-shifting, accommodation, and entrainment as computational models, I aim to realize dialogue systems that adapt their behavior to the relationship with the user and the dialogue context, and thereby build trust with them. I also address the question of how far dialogue systems can grasp meanings implied between the lines, analyzing and enhancing the understanding and reasoning of large language models guided by pragmatic theories.
Such adaptation in dialogue is not carried by words alone: nonverbal behaviors such as gaze, facial expressions, and gestures are equally essential to building relationships. My research therefore spans consistently from observing and analyzing multimodal dialogue in real environments to expressing these behaviors through the bodies of dialogue robots and virtual agents.
Intellectual Property Informatics
Intellectual property practice — covering patents, designs, and beyond — is a highly sophisticated intellectual activity that demands expert knowledge spanning both law and technology, together with strictly controlled language use. Centering my approach on modeling the cognition and decision-making of the professionals who carry out this practice, I develop technologies that support the reading of vast and complex patent documents, assist the writing and drafting of claims that precisely delineate the scope of rights, and sustain international IP practice across languages and jurisdictions. This scope also includes cross-modal information processing for design rights, where drawings and images play a central role.
Going beyond the construction of datasets and models, I pursue real-world deployment through collaborations with industry, aiming to establish natural language processing that is actually used in IP practice.
Collaboration
I welcome research collaborations and inquiries on natural language processing, dialogue systems, and patent informatics. Please feel free to contact me.