Expanding Topics in Speeches Using Topic-Associated Words

Abstract

This paper proposes a system that suggests expanding on related topics to support users in keeping the conversation engaging and flowing smoothly. The system extracts nouns from conversation text to use them as topics. Each topic serves as a search engine keyword, and the suggested words provided by the search engine are related terms for that topic. The relationships between words are visualized as a graph. Additionally, the temporal expansion and changes in topic relations can be intuitively observed as alteration in the structure of the graph.In this paper, we examine whether the graph indicates what content might follow the speaker’s current discourse. If the graph demonstrates the potential to predict the direction of a conversation, it can also help guide the speaker’s discourse in the next developments.