Relevancy Of The Search Result Using A1 Techniques

For any search tools, the main objective is to improve the ratio of the number of hits to number of retrievals. The objective of this project is to develop an intelligent tool for determining the relevancy of the search result, with respect to users' keyword. Northern Light Power
Search engine was used with specific keywords that are related to neural network. Hypertext and computer science domain was the focus of this study. An intelligent model that can categorize search result as relevant or irrelevant to the keyword specified was developed. This software development was divided into two parts, the first part concentrated on software development to count and
categorized keyword in pre-defined format. The second part focus on software development for neural network model with autodetermining capability. Java programming language was used as the programming language to develop the software. Multilayer Perceptron was utilizes as the neural network model implemented in this study. Generic layer and notation of neural network formula were derived from classical model. Prior to the Multilayer Perceptron software development, the data from hypertext keyword counter software was used in "Neural Connection" to confirm the best result that could be acheved. The result from "Neural Connection" has achleved more than 96%. However, the results produced by the developed software decreased by 15%. This may due to the fact that the developed software used non-linear activation function at hidden as well as the output layer.