Agent-Oriented Modeling of Real-Life Discourse

Yueguo Gu

The Chinese Academy of Social Sciences

Beijing 100732

 

Real-life discourses such as a job interview, family photo taking, doctor-patient consultation can be captured through audio- and/or video-recording. The real-life events are thus turned into data deposited in physical audio/video cassettes. The cassette data then can be digitalized into computer-readable media data. The digitalized data are still too “raw” to be useful. Orthographic transcription and annotation are the two dominant methodologies to mine the information from the raw data. The whole process starting from representing real-life events via audio/video recording to orthographic representation assuredly gives the researcher some considerable advantage of manipulating data, but not without paying some heavy prices of losing and even distorting some information which is intrinsically there in the real-life events.

 

Orthographic words and symbols, being static in linear order, are incompetent means to represent real life events, which are produced by human agents. This paper explores an alternative to amend the incompetence. The primary unit of representation is no longer word, but agent, which is an analytic construct used to model a real life target. Take the case of guests wanting to sit down at a banquet table. We can construct three agents the Seat_Arrange_Agent, the Seat_Occupant_Agent, and Person_Agent_Actor to model the real life events. The first agent handles the table layout, the second copes with the rules of who will sit on which chair, and the last models the human decision by individual guests. The triple-agent interaction is visually represented as shown in Figure 1.   

 

 

 

 

 

 

 

 


The bulk of the paper outlines a tentative set of metamodels and visually-based diagrams with which real life discourses in digitalized video format can be modeled for multimodal corpus annotation, speech engineering, human-machine interaction, and AI research.

 

Keywords: agent-oriented modeling, real life discourse, representation

 

Gu, Yueguo, male, 50, Ph. D., a research professor of linguistics, Head of Contemporary Linguistics Department, the Institute of Linguistics, the Chinese Academy of Social Sciences. Research interests include pragmatics, corpus linguistics and discourse analysis. Email: guyueguo@vip.sina.com