Agent Architecture

The architecture of our knowledge-based information agent is shown in Figure . An agent contains three knowledge bases and a framework. The three knowledge bases are: General Knowledge Base (G), Domain Knowledge Base (D), and Site Specific Knowledge Base (S). The framework contains three main functions: Web Access Engine, Information Extraction Engine and Matcher.

  
Figure: The Architecture of Knowledge-Based Information Agents

We detail the three functions as follows:

• The Web Access Engine utilizes site specific knowledge to get access to the Internet and download Web pages. When the information source has a searchable interface, the Web Access Engine needs to use the original user query or the structured query (the IE results from the original query) to interact with the interface.
• The Information Extraction Engine uses general, domain specific and site specific knowledge to extract information from both the user query and Web pages, and to save them as structured data.
• The Matcher matches the structured query with the structured data of Web pages and outputs the results.

An agent shell includes the framework and the general knowledge base. An information agent can be built by adding the domain knowledge base and the site specific knowledge base to the agent shell. An example of an agent shell is given in [8].

As we mentioned in the introduction, this paper focuses on how to use knowledge to extract information from Web page. The next two sections discuss the main components of three knowledge bases and one main function of the framework, the information extraction engine.