Research On Query Expansion Based On Deep Semantics

With the rapid development of I nternet and hi-tech information technology,the information on the In ternet increase s rapidly.Information Retrieval is the main way for user to query and obtain information and the method and means to find information.However,due to the ambiguity of natural language and the incomplete expression of user's query,traditional methods of information retrieval ha ve some limitations,which can not satisfy users' query intention.Therefore,query expansion technology in information retrieval becomes a more and more important research topic.We add new words in the original query so that it can overcome the problem of the natural language with ambiguit y,improve the formulation of que ry intention,and then make the query optimized.Deep semantics is which makes emotional anal ysis,trend anal ysis,geographical anal ysis and relationship anal ysis from massive,redundant,unstructured or structured data with Deep learning or other Natural Language Processing Technology.Making query expansion after building conceptual semantic space between words and anal yzing their deep semantic relati on can mine more relationship between the query words,master users' query intention on the whole.In the passage we propose a method of query expansion based on conceptual semantic space with deep learning.Construct conceptual tree for every query word with deep learning and find a common root node upwards to structure concept semantic spac e in Word Net dictionary.Filter expansion words in source b y co-occurrence information as parameter in order to prevent query drift.Introduce average mutual information and observation window to achieve relevance algorithm,so that measure correlation degree between every two words b y calculating Co-occurrence information.The experimental results show that the proposed method has higher precision and precision.In addition,we obtain deep semantic extended word set of the original query words b y mining the deep semantic relations between them with deep learning,and get statistic extended word set b y statistic method.Then f use semantic extended word set and statistical probabilit y extended word set with Copulas frame to get a better extended word set.Finall y,we screen out the extended words associated with the related information of the query,so as to improve the retrieval performance.