2006 Fiscal Year Final Research Report Summary
Research for Providing an incentive among the Internet users to reach reliable information
| Project/Area Number |
16500080
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Chiba University (2005-2006)
Kyoto University (2004) |
|---|
Principal Investigator |
ARAI Sachiyo Chiba University, Urban Environment Systems, Associate Professor, 工学部, 助教授 (10372575)
|
|---|
| Project Period (FY) |
2004 – 2006
|
| Keywords | Community Extraction / Multiagent Reinforcement Learning / Reward Design / Scale Free Network / Small World / Cooperative Behavior / Network Analysis / Evolving Network |
|---|
| Research Abstract |
We have developed an infrastructure in which "interaction design" and "interoperability" are guaranteed for modeling the multi-agent domain, such as Grid computing, Electric-commerce, and Digital city. Our main interest is collaboration on the web among the agents and human beings. The difficulties are caused by the fact that information sources are distributed widely throughout the WWW environment, and each human has its preference for the particular type of information.
To share and reuse knowledge of others', the agent has to learn about the agents which would provide the reliable and useful information. At the same time, extracting semantics from contents is very critical issue to make information interoperable to be shared and be reused among the heterogeneous agents. The statistical learning techniques were applicable for extracting the semantics and a text classification. While exploring in the scene of the Web, very dynamic and open environment, the agents should be adaptive and
… More
flexible to trawl for appropriate information which satisfy the desire of user's. We have realized adaptive aspects of agents' by reinforcement learning frameworks.
Our research aims at realizing social knowledge through sharing and employing information on the Web. As a first step, we proposed a multi-agent scenarios which enables us to design interaction between users, between users and agents, and among agents, as well. By employing this translator, "interoperability" is guaranteed on the composite Web service, and consequently it becomes reusable to other users. The concepts of "interaction design" and "interoperability" work complementarily for coordinating heterogeneous agents and humans who have different motivations. This overall architecture not only enhances collaboration between users on the Web, but also facilitates user-agent interaction design, and therefore is expected to contribute to Collaborative Commerce and Digital City that will unfold in the Internet in the future. Less
|
