| Cloud computing and big data have promoted the implementation and development of a series of artificial intelligence technologies,enabling people to move further into the intelligent era from the information age.However,due to a series of problems existing in big data itself(unstructured,non-standardized,lack of semantic association,lack of domain background knowledge,low quality,etc.),the value of big data has not been fully explored and utilized.In order to achieve accurate and fine analysis of big data,in addition to relying on domain-related standards,specifications and guidelines,it is more important to rely on the support of domain expertise.At present,although big data-driven machine intelligence has reached or even surpassed the level of human beings in some aspects of perceptual intelligence,its level of cognitive intelligence is still very low.Cognitive ability is the unique ability of human beings.The realization of cognitive intelligence depends on a large amount of common sense knowledge that people have.The research on the modeling method of machine-understandable domain semantic knowledge base and the knowledge service method based on the domain semantic knowledge base has become a research hotspot in the field of artificial intelligence and knowledge engineering.Focusing on the construction of domain semantic knowledge bases,the current research mainly focuses on the general methodology,tool set,knowledge acquisition,knowledge reuse and other aspects.Aiming at the representation of domain knowledge,based on a large number of practical findings,there are still a large number of common semantic mapping problems to be solved.For the research of knowledge service methods,no explicit knowledge service model has been proposed yet,and the current research mainly focuses on the semantic information sharing and semantic integration in the shallow level.Based on domain semantic knowledge bases,there are few studies to provide deep problem solving services for domain applications by supplementing problem solving knowledge.Therefore,based on Ontology,Semantic Web,fuzzy theory and other knowledge modeling technologies and theories,this paper deeply studies the semantic mapping methods in the construction of domain semantic knowledge bases,the reuse methods of open semantic knowledge bases,and the knowledge service model and problem solving methods based on the domain semantic knowledge bases.The main innovations of this paper are summarized as follows:(1)In view of some common semantic mapping problems caused by the polysemy(for example,the same resource name represents both a concept and a property,the same property has multiple domains,the same instance has multiple types,and the same resource is both a type and an instance,etc.)and n-ary relationship(such as uncertain relationships,additional information about relationships,different values of the same relationship,n-ary relationships between multiple instances,etc.)phenomena that are common in domain expert knowledge(referred to as "expert view")and the common semantic mapping problem of access authorization for domain security experts or information owners when providing knowledge services,this paper presents the corresponding semantic mapping solutions,summarizes 10 ontology modeling conventions,and analyzes the semantic mapping results in detail from the aspects of semantic understanding,semantic reasoning and semantic query correctness.The semantic mapping results show that the semantic mapping methods proposed in this paper can ensure the correctness of semantic understanding,semantic inference and semantic query as well as the fine-grained security of access authorization when providing knowledge services.(2)Aiming at the semantic description of fuzzy knowledge in domain expert knowledge,this paper proposes a method of modeling domain fuzzy ontologies and describing domain fuzzy knowledge.Firstly,this method completely follows the ontology design principle,and clarifies that the fuzzy concept is still a concept in ontology,and the fuzzy membership or fuzzy association is still a semantic relationship.Secondly,this method combines fuzzy theory,ontology standard description languages and SWRL to model domain fuzzy ontologies,and can reuse existing OWL(Web Ontology Language)ontology editors and inference engines.In this method,an "OWL 2" individual value restriction constructor is used to characterize the feature functions of fuzzy concepts,an instance value restriction constructor and a class logic intersection constructor are used to blur fuzzy concepts or fuzzy relationships,and an SWRL rule set is used to describe the expression of membership functions.Finally,the method introduces relation classes to express fuzzy membership relations and fuzzy association relations.SWRL rule sets are used to transfer the values of domain elements to fuzzy membership function instances and describe fuzzy membership relations.Automatic calculation of membership degree and automatic construction of fuzzy membership relations can be realized through rule inference calculation and operation.The evaluation results show that the method in this paper is more complete,reasonable and effective.(3)For indirect reuse of the Freebase open semantic knowledge base,this paper identifies and discusses the various obstacles to extract knowledge about a domain directly from a Freebase RDF dump.Based on the complete semantic mapping of Freebase knowledge base and an ontology-based domain semantic knowledge base from three aspects:concepts,knowledge representation models and semantic description components,this paper proposes a method called "EdokFred".On the one hand,the method can pre-process the Freebase RDF dump package,and ensure the integrity of the knowledge of each subject matter domain while reducing the scale of the dump package as much as possible.On the other hand,it is possible to quickly,accurately and completely extract the ontology schema and instance data of a certain domain from the processed dump package,and convert it into a form described by the ontology standard languages.The evaluation results show that the proposed method is superior to the existing methods in terms of processing results,accuracy,integrity,processing performance and reusability.(4)In order to provide knowledge services based on domain semantic knowledge bases,this paper proposes a knowledge service model based on domain semantic knowledge bases.Taking the application requirements of the intervention program recommendation for chronic patients and the symptom-based aided diagnosis of common diseases as an example,a semantic knowledge base in the domain of health care was constructed by using the domain semantic knowledge base construction methods proposed in this paper.The diet and exercise recommendation models for patients with chronic diseases were constructed with the participation of experts as the complementary problem solving knowledge.The evaluation results show that the semantic knowledge base of health and medical domain constructed in this paper can meet the requirements of the recommendation models,and the recommendation results are basically in line with the patients' physical condition and intervention target requirements.In view of the problems and deficiencies of the "3R method",this paper proposes an "IMP3R method".With the proposed aided diagnosis algorithms for common diseases as the supplementary problem solving knowledge,the IMP3R method in this paper is higher than the 3R method for the Top-1 and Top-3 hit rates of 6 common diseases randomly selected.Compared with naive Bayes classification and decision tree classification methods,the IMP3R method in this paper can avoid the "cold start" problem and can quickly support the aided diagnosis of a large number of common diseases. |
PDF Full Text Request