Image re-ranking semantic search engine : Reinforcement learning methodology

Information retrieval and web search present a Challenging Question to researches. Today users urge for accurate and precise hands on information from Search Machine. Interpreting of user query goal is major challenge in past and present. Numerous algorithms and Frameworks have be proposed, but fail to incorporate user aims, as query without proper intent processing retrieves irrelevant information pattern discovery has ability to solve in limitations of keyword and image disambiguates with phrase learning ie, pattern discovery. Today's search machines are based on ranking model eliminating Boolean retrieval constraint and boosting natural language use. Even though word sense and concept extraction is major challenge which comes up with keywords. Information can be presented in better way with image presentation, which is been used in news portals to communicate fastly happing news and social websites instagram Facebook, flicker .user purchase goods by sighting product images on flipkart. So today uses have sifted their approach from text based information to image based, which has given rise to research domain of image information retrieval (IIR) but large number of image attributes also give rise to Image classification ambiguity. Relevance is major factor that influence information retrieval system performance with impact precision and recall. Relevance re-ranking is methodology opted in to retrieve most optimized relevant results eliminating non-relevant. Large amount of image with associated word annotations are present on different web portals. In this research we build a semantic search engine which selects network design pattern and integrate reinformant learning approach (Agent based learning) that help in selecting information from various networks and help in network structuring with WAIR (Web Agents for Information Retrieval) Architecture at core. Agent helping in retrieving precise objects from different portals and linking them. A optimized procedure E-SimRank is been implemented to count in link semantic in network and content based knowledge learning for reinforcing better results. Performance evaluation show that proposed architecture and algorithm design present faster and relevance result. A image based recommendation system is our research outcome which contributes to image retrieval domain. The research work is been developed by studying 24 core vital articles on image retrieval and find research scope with major challenges which have common ground and need to be addressed. The found Research Analysis Query (RAQ) help in directing to study better techniques to overcome problem. Our research innovation is reinforment learning algorithm agent based system development. Existing state of art of present algorithms have been optimized with this innovation integration. Future scope of research lies to image to image base retrieval or video recommendation system.