Abstract
Keyword-search is the prevalent method for finding information on the Web today. However more expressive methods to acknowledge the intriguing characteristics of today’s user/content combination on the Web are required. Based on this premise, this paper sets forth to investigate a very expressive method called relevance feedback, on which the user judges the relevance of pages and continuously redirect the search toward the assumed user-preferred pages. This proposal incorporates relevance feedback into a completely functioning search engine to improve the functioning keyword-searching mechanism. We investigate a number of web page features, which could be pursued for ranking pages according to relevance feedback. By applying supervised learning algorithms, we aim at determining how those features should be weighed for the best outcome in ranking and thus engineered for relevance feedback.
Recommended Citation
Durao, Frederico Araujo; Dias de Almeida, João Paulo; Santos Peixoto, Daniel; de Souza, Paulo Roberto; Schjønning, Carsten; and Bech Rasmussen, René, "Exploiting Web Features for Relevance Feedback" (2019). AMCIS 2019 Proceedings. 1.
https://aisel.aisnet.org/amcis2019/ai_semantic_for_intelligent_info_systems/ai_semantic_for_intelligent_info_systems/1
Exploiting Web Features for Relevance Feedback
Keyword-search is the prevalent method for finding information on the Web today. However more expressive methods to acknowledge the intriguing characteristics of today’s user/content combination on the Web are required. Based on this premise, this paper sets forth to investigate a very expressive method called relevance feedback, on which the user judges the relevance of pages and continuously redirect the search toward the assumed user-preferred pages. This proposal incorporates relevance feedback into a completely functioning search engine to improve the functioning keyword-searching mechanism. We investigate a number of web page features, which could be pursued for ranking pages according to relevance feedback. By applying supervised learning algorithms, we aim at determining how those features should be weighed for the best outcome in ranking and thus engineered for relevance feedback.