Research And Application Of Knowledge Question Answering System In Education Based On Knowledge Graph

With the development of the artificial intelligence industry,the demand for talents in this field is growing rapidly.And more and more students will engage in jobs related to artificial intelligence.However,colleges and universities are in the initial stage of cultivating talents for artificial intelligence industry,and they are also lack of the teaching environment and platform considering talent demand.Meanwhile,in the era of information explosion,students who are working in AI field need to search related knowledge or technologies or requirements for applications from multiple platforms.But the infonnation returned from those platforms is too massive,and it is also time-consuming and high-labor-cost to extract useful information from those results,such as relevant knowledge and the recruitment requirements of the company.In addition,data from multiple platforms would usually have heterogeneous format.That makes it difficult for machine to integrate and utilize network data from multi-source,not to mention to deeply extract the semantic information.Knowledge graph technology is used to quickly describe domain concepts and relationships between entities.It is one of the methods to solve integration and analysis problem of multi-source heterogeneous data.This is a valuable research topic,and it has many important applications.This paper provides a knowledge graph in education domain,basing on the entity recognition algorithm and the word co-occurrence extraction algorithm.We also create a knowledge question-answering system,which could provide students with knowledge services basing on natural language interaction.The main work of this thesis are as follows:1.In order to solve the problem of heterogeneous data from multi-sources,I define an entity model connecting the needs of enterprise users with academic resources,in which various entities are associated with domain concepts.2.In view of the fact that various entities in enterprise job data have undefined boundaries,multiple representations and unclear text features,this paper proposes an entity recognition algorithm that combines convolutional neural networks and two-way long-term and short-term memory networks.The experimental results show that the proposed method can better identify the required entities and improve the accuracy and recall rate of entity recognition.3.It is difficult for a number of job entities to directly describe the skills required to master a field and the mastery of different skills.In order to solve the problem,this paper proposes a word co-occurrence extraction algorithm based on multi-factor mixture,which combines word co-occurrence relationship,degree word meaning similarity and subjective understanding to calculation weight.The capability model extracts the comprehensive recruitment requirements of each core field of artificial intelligence.The experimental results show that the field ability model extracted by this method has good guidance and reference significance for students to develop learning content and understand domain knowledge.Based on the above algorithm,this paper constructs an educational knowledge graph around the core field of artificial intelligence,and builds a domain-based question and answer system based on knowledge base.It can use natural language to talk with students in real time,helping students to quickly grasp the domain overview.Meanwhile,students can formulate learning goals and learning content based on the company's needs for talent,to reduce the cost of information search,thereby improving learning efficiency,accelerating the emergence of outstanding talents in the field of artificial intelligence,and achieving career goals.