作者: Antti Kerminen , Kristiina Jokinen , Mauri Kaipainen , Kari Kanto , Topi Hurtig
DOI:
关键词:
摘要: In this paper, we present our approach to dialogue management in the spoken system that is being developed within project Interact. Compared traditional approaches, manager will support system’s adaptivity and flexibility with help of two design decisions: an agent-based architecture use neural network models. Our experiments focus on word-based act recognition using LVQ classification algorithm a corpus information-seeking dialogues, compare results simple bag-of-words approach. We also report studies clustering input data into necessary meaningful categories self-organizing maps.