BRAZILIAN HEALTH-RELATED CONTENT WEB SEARCH PORTAL Presentation on a Method for its Development and Preliminary Results

The increase in the amount of available information on the world wide web is inexorable, which, on one hand, provides the web user with more information. On the other hand, however, web searches become increasingly more difficult to handle due to the increasing number of retrieved documents. The present study is a proposal of development for a Brazilian search portal specific for health-related content. The aim of such development is to provide web users, mainly the non-specialist ones, with the largest number possible of web pages relevant to their search terms and inferred search intentions. The proposed search portal integrates web mining-based filters and a decision-making support tool. The preliminary study results show that among the algorithms tested to incorporate a filter module specific for health-related content artificial neural networks, logistic regression and nearest neighbor clustering (NNC) -, the application of NNC resulted in the automated web health-related content classifier with the best performance for sensitivity and specificity 0.92 and 1.00 respectively.Univ Fed Sao Paulo, Dept Hlth Informat, Sao Paulo, BrazilUniv Fed Sao Paulo, Dept Hlth Informat, Sao Paulo, BrazilWeb of Scienc