Development of Multi-Agent-based Intelligent Information Retrieval System

在本研究中，我们首先提出了基于agent的点对点信息检索（ACP2P）方法，并进行了各种实验。ACP2P方法使用代理社区来管理和查找用户感兴趣的信息。代理作为其用户的委托，并通过与其他代理通信来搜索用户想要的信息。代理之间的通信是在对等计算体系结构中进行的。检索与用户查询相关的信息使用内容文件执行，内容文件由原始文档和检索文档以及两个历史记录组成：查询/检索文档历史记录和查询/发送方代理历史记录。利用历史具有协同过滤效果，逐渐创建虚拟代理社区，其中具有相同兴趣的代理聚集在一起。我们的假设是，虚拟代理社区减少了执行搜索所需的通信负载。当一个代理接收到更多的查询时，就会获得更多的新知识链接。从这种行为中，代理人的“给予和接受”（或积极反馈）效应似乎出现了。其次，提出了个人语义Web信息检索系统的概念体系结构。它结合了语义Web、Web服务、P2P和多代理技术，不仅可以实现Web资源的精确定位，还可以实现通过Web内容和Web服务交付的Web资源的自动或半自动集成。在这个体系结构中，一致地考虑了与信息检索的整个生命周期有关的语义问题，并且无缝地实现了Web内容和Web服务的集成。该体系结构由三个主要组件组成：消费者、提供者和中介。所有的提供者和消费者都被构造为语义上的“MyPortal”，它为与用户相关的所有信息提供了一个网关。每个提供者在我们称为WSCD （Web站点功能描述）的内容中描述其功能，当需要进行Web搜索时，每个使用者将根据用户需求提交相关查询。中介由分配给消费者和提供者的代理组成，使用基于代理的P2P信息检索（ACP2P）方法实现语义myportal之间的信息共享。一些初步的实验结果表明了ACP2P方法的有效性，以及两个查询/响应历史在查找新信息源和减少通信负载方面的有用性。最后，提出了支持用户网页浏览活动的主动浏览系统。系统观察用户的网络浏览行为，了解用户的方式或偏好，并通过推荐来节省用户到达所需页面的劳动。该系统不仅推荐用户过去浏览过的页面，而且通过协同过滤技术从其他用户中收集用户感兴趣的新页面。该系统是基于Multi-Agent KODAMA框架实现的。一些初步的实验结果说明了我们方法的有效性。少

In this research, we first proposed the agent-community-based peer-to-peer information retrieval (ACP2P) method and carried out various experiments. The ACP2P method uses agent communities to manage and look up information of interest to users. An agent works as a delegate of its user and searches for information that the user wants by communicating with other agents. The communication between agents is carried out in a peer-to-peer computing architecture. Retrieving information relevant to a user query is performed with content files which consist of original and retrieved documents, and two histories : a query/retrieved document history and a query/sender agent history. Making use of the histories have a collaborative filtering effect, which gradually creates virtual agent communities, where agents with the same interests stay together. Our hypothesis is that a virtual agent community reduces communication loads necessary to perform a search. As an agent receives more queries, then m … More ore links to new knowledge are acquired. From this behavior, a "give and take"(or positive feedback) effect for agents seems to emerge.Secondly, we proposed a conceptual architecture for a personal semantic Web information retrieval system. It incorporates semantic Web, Web service, P2P and multi-agent technologies to enable not only precise location of Web resources but also the automatic or semi-automatic integration of Web resources delivered through Web contents and Web services. In this architecture, the semantic issues concerning the whole lifecycle of information retrieval were considered consistently and the integration of Web contents and Web services is enabled seamlessly. The architecture consists of three main components : consumer, provider and mediator. All providers and consumers are constructed as semantic "MyPortal" which provides a gateway to all the information relevant to a user. Each provider describes its capabilities in what we call a WSCD (Web site capability description), and each consumer will submit relevant queries based on user requirements when a Web search is necessary. The mediator is composed of agents assigned to the consumer and providers using an Agent-Community-based P2P information retrieval (ACP2P) method to fulfill the information sharing among semantic MyPortals. Some of preliminary experimental results showed the efficiency of the ACP2P method and the usefulness of two query/response histories for looking up new information sources and for reducing communication loads.Lastly, we proposed the Active Browsing system which supports user's Web browsing activity. The system observes user's Web browsing behaviors, learns his/her ways or preferences and tries to save his/her labor for reaching the pages he/she wants by recommending them. The system recommends not only the pages the user browsed in the past, but also the pages which are new and interesting to him/her by collecting from other users with a collaborative-filtering technique. The system is implemented based on Multi-Agent KODAMA framework. Some of preliminary experimental results illustrated the validity of our approach. Less