Design And Implementation Of Customer Service Q & A System Based On Domain Ontology

At present domestic e-commerce platform mainly rely on artificial customer services solving users’ good pre-sales and after-sales questions. In order to reduce the workload of customer service and improve the efficiency of the system, this thesis build the knowledge base combined with user common problem and domain knowledge, realize the customer service question answering system based on domain ontology. At present, in question answering system, domain ontology is mainly used for the concepts taxonomy of target domain, lack semantic understanding for user’s questions. This thesis focus on the study and improve the description of property and property similarity calculation methods. Research work mainly focuses on the following aspects:Firstly, study the expression of domain ontology concepts and their relationships, build the knowledge of domain ontology about mother-to-child. In the process of ontology construction, not only expresses the relationship between the concept of classification, but also emphasizes the description of the entity’s property. Take advantage of sentence dependency parser analysis, map the subject-predicate-object in the natural language to the subject-predicate-object in ontology.Secondly, study the concept similarity calculation method in domain ontology, put forward improvement of concept similarity calculation method in domain ontology. Previous calculation of the ontology concept similarity, it mentioned "concept" as the classes or individuals in the ontology. Based on domain ontology building, this thesis designed and implemented similarity calculation based on ontology which could calculate the similarity of classes, instances, and properties. Realize the concept similarity calculation between the concepts’ non-taxonomy relationship.Finally, according to the concept similarity calculation methods above, propose a method of sentence extension methods based on ontology. The method according to the taxonomy relations of the classes to delete redundant information of a sentence, and according to the non-taxonomy relations between classes expand the potential information. Finally combining VSM algorithm, calculate sentences similarity, build a question answering system about mother-to-child knowledge.The algorithm model has been successfully applied in HelloOWN--Qingdao Haier Software intelligent customer service question answering system, has achieved good performance in semantic analysis after use this mother-to-child ontology.