Exhibition Program

Science of Communication and Computation


Robot understands events in your story

- Chat-oriented dialogue system based on event understanding -


The proposed chat-oriented dialogue system can make users think the system understands the user’s utterances. By understanding the user’s utterances as an event structure (a group of time, place, person, etc.), we achieve a chat-oriented dialogue system that can sympathize and delve into topics during a chat. To understand a user’s events from the user’s utterances, a system must understand various words/phrases in user utterances. To tackle this problem, we focus on general words and phrases that are familiar in a chatting situation but difficult to extract by conventional methods. Using this technology, systems can extract a user’s utterances by organizing the extracted information. In the future, we aim to foster a world where humans can converse with systems like humans with mutual understanding by grounding the extracted information to the system and external knowledge.


  • [1] H. Narimatsu, H. Sugiyama, M. Mizukami, “Detecting Location-Indicating Phrases in User Utterances for Chat-Oriented Dialogue Systems,” in Proc. The Fourth Linguistic and Cognitive Approaches to Dialog Agents Workshop (LACATODA), 2018.
  • [2] H. Sugiyama, H. Narimatsu, M. Mizukami, T. Arimoto, “Empirical study on domain-specific conversational dialogue system based on context-aware utterance understanding and generation,” JSAI SIG-SLUD, 2018. (in Japanese)
  • [3] M. Mizukami, H. Sugiyama, H. Narimatsu, “Event Data Collection for Recent Personal Questions,” in Proc. LACATODA, 2018.




Hiromi Narimatsu, Interaction Research Group, Innovative Communication Laboratory