Query Conceptual Graph Guided Web Search Result Analysis

In modern society, web becomes an important information source and search engine is the main entrance to web resources. The performance of search engine decides whether we could accurately reach target information in a short time. Popular search engines are usually based on Boole model, vector space model, or language model which take query and web page as sets of discrete word and only consider word co-occurrence. These models cut off relations between words and ignore latent semantic information in text, which brings high noise and leads to low precision. Aimed at handling above problems, researchers propose several kinds of semantic based retrieval models.Intension Semantic Conceptual Graph model tries to construct CG for text, extracts its concepts and rebuilds conceptual relations. Since CG can better represent semantic information of text, it could be used in semantic retrieval as semantic description method.Our work includes four parts. In the first part, we propose a CG based framework of web search result analysis and manually process several instances under this framework.In the second part, we propose a method to transform the snippet into reduced snippet CG. This method avoids web page indexing and focuses on snippet indexing. It is highly automatic and independent of complete syntax parsing.In the third part, we analyze wrong web search results and sum up common error types. Then we propose two methods to filter wrong items in web search results. Both methods are based on query CG and reduced snippet CG. Our experiment shows that combination of the two methods can effectively increase retrieval precision.In the fourth part, we try to modify both filtering methods and apply them to web search result sorting, which is a worthy exploration of query CG guided web search result sorting.We try to analyze web search result under the guidance of query CG and hope to provide some threads for the research of CG based semantic retrieval model.