Knowledge Map Construction And Its Core Feature Extraction For Multivariate Space Big Data

With the development of the earth's information and the Web,the data in the multidimensional earth space,which mainly composed of geospatial space and Web space,tend to be volum,velocity,variety and value.Such complicated and large-scale multidimensional geospatial data should have provided more information to users.However,due to the lack of effective knowledge mining,organization and retrieval methods,this vast amount of information make users always be lost on the knowledge.The demand of users for efficient knowledge organization,mining and retrieval methods are more and more urgent.In order to provide users with high-quality knowledge services in large-scale,sparsely related and heterogeneous data at multiple spaces,we face the following challenges: 1)The scale of text data in Web is large and its value is sparse,so it is hard to carry out knowledge mining.2)It is hard to represent the knowledge in Diversified heterogeneous and dynamic data at multivariate space.3)It is difficult for knowledge retrieval when user needs vary from person to person and changeable.To deal with the above challenges,this paper proposes a Knowledge Map Construction and Its Core Feature Extraction Method for Multivariate Big Data.The goal of our method is to discover the compact semantic knowledge from the massive multivariate space big data and organize it with the knowledge map,finally to provide users with accurate and rapid knowledge retrieval service.For this purpose,the main reaearch work of this paper is as below:(1)Based on the model of textual power series representation,a core sentence selection method based on entity in Internet space text data is proposed.This method excavated the closely related keyword association patterns in textual data and used it as the basic unit of sentence semantic representation.This representation makes the semantic rich sentence is given a higher semantic weight,to ensure the semantic richness of extracted sentences.The method also introduces the semantic decay function to control the semantic redundancy of core sentences so as to ensure the simplicity and novelty of the extracted sentence set information.(2)Utilizing the complementary features of multivariate spatial information,a method to construct the geographic entity knowledge map in multivariate spatial data is proposed.This method excludes the attribute relations of geographic entities in multivariate space to improve the recall rate of entity relations in the map.The sentence structure is analysed by the method of dependency syntactic analysis,and the heuristic rules of "entity pair" and "entity pair" feature sequence mining are used to extract the entity relationship from the sentence and improve the accuracy of entity relation extraction(3)Based on the idea of context extension and the extension of context,this paper proposes a method to extract the features of related entities from the map of geographical entities.This method uses the background knowledge to expand the context of the entity,reduces the extrapolation of the entity-independent features in the knowledge map through the mutual constraints between contextual entities,making the entity feature set can be dynamically generated with the change of background knowledge,with the feature extraction of adaptiveThis paper focuses on established the theory and method of geographic entity association semantic mining,entity relation extraction,entity feature extraction.The research promotes the development of multivariate spatial data fusion of geographic information knowledge map building method.The research results can be applied to knowledge recommendation system,knowledge retrieval system,intelligent questionanswer system and so on.