An intelligent WWW agent for similarity-based searching

The paper describes the development of a WWW agent that uses similarity-based methods to search the Internet. The Internet Information Agent (IIA) works by analysing a sample of the type of text that is known to be of interest to the user. It then extracts a number of linguistic features and stores these as a feature vector that is used to describe the content of the document. This data is then used as input to a range of similarity metrics that allow the agent to compare new texts with the original and thereby acquire “more of the same”. The agent´s strengths lie in its use of a range of similarity metrics that are known to perform well over a wide variety of input. The agent has been tested across a range of input data and evaluated against a number of criteria. The results of this evaluation are described and the prospects for the ongoing development of the agent are discussed