Event Sequence Model for Semantic Analysis of Time and Location in Dialogue System

Event Sequence Model for Semantic Analysis of Time and Location in Dialogue System Yasuhiro Noguchi Faculty of Informatics Shizuoka University noguchi@inf.shizuoka.ac.jp Satoru Kogure Graduate School of Informatics Shizuoka University Makoto Kondo Graduate School of Informatics Shizuoka University Ichiro Kobayashi Graduate School of Humanities and Sciences, Ochanomizu University Hideki Asoh Intelligent Systems Research Institute, AIST Akira Takagi NLP Research Laboratory Tatsuhiro Konishi Graduate School of Informatics Shizuoka University Yukihiro Itoh Shizuoka University Abstract It is important for a natural language dia- logue system to interpret relations among event concepts appearing in a dialogue. The more complex a dialog becomes, the more essential it becomes for a natural language dialogue system to perform this kind of in- terpretation. Traditionally, many studies have focused on this problem. Some dia- logue systems supported such semantic analysis by using rules and/or models de- signed for particular scenes involving spe- cific type of dialogue and/or specific problem solving. However, these frame- works require system developers to recon- struct those rules/models even if a slight change is added to the targeted scene. In many cases, their rules/models heavily de- pend on specific type of dialogue/problem solving, and they do not have high reusabil- ity and modularity. Since those rules/models have scene-depending design, they cannot be used to incrementally construct a bigger rule or model. In this research, we focus on a set of event concepts which are usually expected to occur sequentially. In a dialogue, a spoken event concept enables the listeners to guess a sequence of events. The sequence may sometimes be logically inferred, and it may be understood based on general com- mon sense. We believe that a concept model of sequential events can be designed for each bigger event concept that consists of a series of smaller events. Using the sequentiality of the events in the model, a dialogue system can analyze time and loca- tion of each event in a dialogue. In this pa- per, we design a structure of the event se- quence model and propose a framework for analyzing time and location of event con- cepts appearing in a dialogue. We imple- mented this framework in a dialogue system, and designed some event sequence models. We confirmed that this system could ana- lyze time and location of sequential events without scene-depending rules. 1 Introduction Natural language understanding by a NLP sys- tem requires more than generating semantic rep- resentations corresponding to the user's input sentence and adding them into the dialogue con- text information in the system. The system is also required to interpret the semantic representations generated from the user's input sentences by comparing them with the dialogue context, the situation surrounding the user and the system, and common sense knowledge. Through these analyses, the user’s input is correctly/restrictively understood beyond what is explicitly uttered. In our previous research, we focused on the difficulty for a natural language dialogue system arising from synonymous expressions. The same semantic contents can be conveyed by different expressions with different set of words and …

Content truncated to 500 words