On selection of topics that users are interested in but are not familiar with
Yuya Sakai, Mitsuharu Matsumoto · 2022 International Power Electronics Conference (IPEC-Himeji 2022- ECCE Asia) · 2022
In this research, we constructed a topic selection system in order to provide “topics that users are interested in and are not familiar with” from the dialogue system side. To achieve this goal, we used a concept distance to provide topics that the user is not familiar with, and improved the accuracy by removing those with a short conceptual distance. Many words in the conversation are not included in the thesaurus dictionary. We constructed a conceptual distance measurement model using machine learning so that the conceptual distance can be measured even for words that are not included in the thesaurus dictionary. As a result of the subject experiment, it was confirmed that the proposal system can be used to provide topics that the user is interested in and is not familiar with.