Using the genetic algorithm to optimize Web search: Lessons from biology

Abstract

Searching for information on the Web is a relatively inefficient process. My goal is to develop a method that optimizes web search queries without user intervention. Developing intelligent ways to automate this process includes the development of algorithms that automatically manipulate the use of keywords to produce the desired output. Genetic algorithms (GA) provide a potentially useful approach in this area. However, these approaches have not fully exploited the biological concepts associated with genetic reproduction and evolution. I hypothesize that an approach that uses GA but modifies it to include the biological concepts of structural and regulatory gene types and the use of a combination of deletion operator and silent genes will improve GA performance in optimizing Web search. In this paper, I describe this approach and its implementation in simulations of Web search tasks using three popular Web search engines (Google, Yahoo and Netscape). The results of this implementation are presented and are compared to the performance of a similar, but unmodified GA in the same tasks. (Abstract shortened by UMI.)