Content Management System Mediated by Conversational Agent

Content Management System Mediated by Conversational Agent

by Hidekazu Kubota

Abstract

The purpose of this research is to develop a content management system for personal content creation. The creation of electronic content, such as e-mails, electorical documents and web pages, is essential work for human intellectual life at present. The size of electronic content is huge because it is suitable for accumulation, consequently people needs computational support for content management. However, it is impossible to accumulate personal content in resusable form because personal content involves subjective and miscellaneous information. Moreover, there is no method of mediating among content creators and users to stimulate content creating process.

This thesis introduces the solution for these problems by using a content management system mediated by a conversational agent. Firstly, this thesis proposes the method of developing EgoChatII that is a content management system using virtualized-egos. A virtualized-ego is the agent that is able to publish content on behalf of an actual person. By using virtualized-egos, people can circulate their content anytime and anyplace. EgoChatII enables people to accumulate their personal content into microcontent that is resusable form and can represent subjective and miscellaneous personal information. EgoChatII also enables people to circulate their content conversationally by using the method of generating multi-agent conversation from microcontent. The feasibility of near-natural conversation of virtualized-egos is certificated by experimentation.

Secondly, the method of developing POC caster is proposed. POC caster is a broadcasting agent system using conversational representation. EgoChatII left the problem that microcontent is too small to build large content, however POC caster solve this problem by using a POC card that consists of a nugget of text and image. POC caster provides conversational representation that is transformed from text information in a POC card. Conversational representation is a dialogic representation between a main caster and an announcer. The understandability of the conversational representation is certificated by psychological experiment. Besides, POC caster is transplanted to a settop box that is connected with TV and evaluated empirically by one year experiment.

Finally, the knowledge channel model that organizes a personal content management system is proposed. The knowledge channel model is materialized by developing EgoChatIII system, a program table system and the channel policy. EgoChatIII enables a content creator to get feedback from users by using the virtualized-ego that can act as the creator’s proxy. The channel policy is the policy document for content management. It enables people to declare the intention of circulating content. The program table system generates overlook of huge content from channel policies. This personal content management system is used practically.

This thesis concludes the realization of the personal content management system by the above results.