Research On Location Semantic Calculation Method Of Place Name And Address

With the rapid development of artificial intelligence technology,person's lifestyle is more and more closely connected with the Internet,which further promotes the research of map navigation,peripheral information search,location query and other technologies.Among them,location information service has a more important position.Due to the openness and freedom of Internet communication,the location information obtained from the Internet is lack and redundant,resulting in ambiguity of address semantics,which seriously affects the accuracy of location services.Therefore,in order to integrate the location information into the computer system more perfectly,it is particularly important to study the semantic calculation method of place names and addresses.In this thesis,the address text obtained from the Internet through web crawling method is taken as the research object,and the address location information mainly includes the administrative division and the road street number,so the problems of these two parts are mainly solved.1.Address administrative division information has the problem of location semantic ambiguity caused by missing feature word and repeated naming.Therefore,this thesis proposes a method of calculating administrative division based on credibility model,and the evidence credibility and the conclusion credibility in the credibility model is corresponded to address factor and administrative divisions obtained through the maximum moving window match algorithm.Combining with uncertain reasoning method,the administrative division conclusion with the highest credible value is obtained as the identification result.2.The road street number is mainly an ambiguity problem caused by the redundancy of feature words.To this end,based on the accurate acquisition of address administrative division,this thesis proposes a more complete method for semantic calculation of address location.By counting the word frequency and the adjacent word frequency of feature word in the corpus,the transition probability matrix of feature word is constructed,and the location information of place names and addresses is accurately partitioned by combining the maximum joint probability algorithm of strings.The address information in the text is taken as a test sample in this thesis,and compared with the traditional address matching algorithm,the mainstream Chinese word segmentation algorithm and the algorithm based on empirically constructing the transition probability matrix.The experimental results show that the proposed method has higher accuracy.While obtaining the location semantic information of the address,it can normalize the address structure and make the geographical location information service better applied to the society.