The Research On Key Technology Of Semantic Search Engine In Semantic Web

Semantic web is the trend of future network. Document in semantic web contains enough information to process data semantically. Focused Search Engine which is particular work in semantic web environment takes full advantage of semantic information. It enables search engine to parse document in the form of semantic annotation or ontology which makes accurate search within specific field become possible. It also improves the technology of Internet and makes sense to prevalence and development of semantic web.Focused crawler, Ontology matching and semantic indexing constitute the core function of whole search engine as well as search interface. The research key point of this paper is followings. (1) New method for the similarity computation. This new method takes more into consideration than ever which is multiple connections between concepts and property also make contribute to the similarity of concept it belong to. (2) Algorithm for focused crawler to make decisions during crawling. We take similarity computation for the document to enhance Reward Function of Q-Learning, and user data to make judgments on parameters. Both of them make effort to performance of related semantic document collection. (3) Ontology mapping based on multiple methods of similarity measure. Not only from linguistic and structural similarity between two concepts but also from instance similarity as the result of ontology matching. (4)Reasoning for semantic document and hierarchy indexing. The aim of indexing is to set up hierarchy and classified index for document which is crawled by crawler. Semantic index includes different levels which is document, clustering, path and triples. It could satisfy different need from user query. It could also predict user's search interest by compute similarity between information and improve search efficiency.This paper makes creative exploration and contribution on following points:(1)This paper feature semantic web and propose limitation of current semantic search engine. Based on analysis, this paper propose vertical semantic engine framework which is under semantic web (FSTSE for short). We describe workflows and functions of each part. FSTSE is the fundament of all research in this paper.(2)Based on FSTSE, we describe and analyze algorithm for link prediction of focused semantic crawler. Semantic document in current web is scattered spread and has a large quantity. The ability of link prediction and content analyze is very important for semantic focused crawler. In this paper, we present semantic document in form of graph structure which could give a clear and accurate description of semantic document. Based on this from of representation of document, we propose QBLP algorithm. It combines reinforcement learning and Bayes classifier to make a prediction when crawler has to make decisions between paths. QBLP is mainly responsible for providing knowledge to pick up best link to crawler through accumulation knowledge from both semantic link and document. Experiment shows QBLP succeed in promoting performance of semantic focused crawler.(3)In order to resolve problem of ontology heterogeneous, we propose a new method of ontology matching which take similarity from different aspect into consideration. Except for linguistic and structural similarity that is common use, we also put emphasis on instance similarity computation. Further more, a new method of clustering for semantic distance has been brought up, as well as a standard to measure cluster which named"purity of information entropy". We use it to analyze similarity between two concepts which possess a set of instances. According to the statistics, this algorithm shows a promising result on improving efficiency of ontology mapping.(4) The response efficiency of semantic query is always one of the obstacles in semantic web. To tackle this issue, we make a comprehensive parse on semantic information and set up structural hierarchy index which is to meet the different purposes or different level of requirements from user's query. Structural hierarchy index analyze different features from different objects. In this way, user's action is directed step by step according to the index until real purpose is exposed. A serial of experiments are proceeded to test both function and performance.(5) Based on FSTSE, we have already set up a semantic search engine, named Sniper, in our lab with purpose of research. It is constituted of semantic focused crawler with dynamic strategy of decision make, ontology matching with multiple ontology elements similarity measure and hierarchy structural index.Sniper realizes all algorithms stated above. It is capable of giving quick response of user's semantic query. Experiments we make proved that Sniper gives a high performance according to statistics.This paper aims to resolve the problem of information integrated under the surrounding of semantic web. And based on that, it makes a realization the whole process of information retrieval, merging, organization and use. It is an application of research on information integration...