The Methodology For Automatic Construction Of Educational Ontology

Educational informatization and smart education are the development trend of education in the future.An important feature of educational informatization is educational intelligence,that is with the help of the Internet,to provide high-quality personalized intelligent learning environment,meet the personalized needs of learners,teachers and managers,and realize the effective sharing and reuse of learning resources.In short,the learning process is a knowledge acquisition process.Smart education involves an environment of knowledge dissemination and knowledge acquisition.It can be said that smart education infrastructure is a knowledge platform,so the identification,formalization,organization and sustainable use of knowledge and learning components become very important.To some extent,the problem of smart education is transformed into knowledge management issues.Therefore,how to make the teaching materials,teaching plans and teaching process more knowledgeable is a new challenge.At present,the urgent demand of smart education is how to introduce semantic intelligence into the knowledge of teaching materials,teaching programs and teaching processes.Ontology,as a formal and explicit specification of the shared conceptualization,can meet the above requirements of educational informatization and intelligence.And the most important characteristic of ontology is knowledge sharing and reuse.Therefore,this paper studies the ontology-based education(Educational ontology).Educational ontology is an important tool and means to realize educational informatization and smart education,but how to build an euducational ontology?In traditional ontology learning,ontology is mostly built by hand.It is not only time-consuming and error prone,but also boring and difficult to expand.With the boom of text data and web-based resources,manual construction has been unable to satisfy demand.There-fore,how to construct an educational ontology automatically is becoming a hot topic.This thesis aims to build a domain ontology by combining the advantages of web-based education with ontology,and proposes an automatic construction framework for educational ontology.At the same time,some key technologies involved in the framework are studied,including knowledge extraction,coreference resolution,automatic ontology expansion and so on.The main contributions of this paper are as follows:�In view of the scarcity of automatic construction and expansion methods of educational ontology,an educational ontology is designed and an automatic construction and expansion framework of educational ontology is proposed.First of all,we propose an education domain ontology which includes three sub categories(course materials,guidance materials,personal knowledge).Then,for this ontology,we propose an ontology automatic construction framework,which mainly applies several technologies in the field of natural language processing to auto-matically extract knowledge from text data to form an original ontology.Finally,considering the ecological expansion of ontology,based on the original ontology,an automatic expansion framework of ontology is proposed.The above methods have been realized in physics teaching materials,and a physics education ontology has been successfully constructed.�A joint extraction method of entities and their relations for multi-head problem is proposed.The main purpose of our method is to solve the multi-head problem which is insurmountable in the existing end-to-end joint extraction model.Firstly,our method introduces a novel dual tagging pattern,and then based on this tagging pattern,an entity-propagation model is proposed,which transforms the extraction task into the prediction problem of dual tagging sequences.The results of the experiment on the public and physics education ontology datasets show that our method can effectively deal with the multi-head problem.�A joint extraction method of entities and their relations for multi-relation prob-lem is proposed.This method mainly uses formal concept analysis(FCA)to cluster the relations with semantic relevance,which is similar to the concepts in ontology,and is responsible for clustering the multiple relations(1-N-1)among entities into one-to-one situations.In this way,we can use the end-to-end joint extrac-tion model to extract knowledge.The results of the experiment on the public and physics education ontology datasets show that our method can effectively deal with multi-relation problem,moreover can explore the potential relationship between entities.�A mention-pair coreference resolution model based on neural network is proposed,which is mainly to solve the issues of low efficiency and global inconsistency.Because many existing methods consider all text spans of the whole document,resulting in a high time complexity of the model(O(T~4)),so we maps the coreference resolution into a graph,and apply the graph convolution neural network to identify whether the mention pairs are coreferent.The experimental results show that our method is efficient and effective in the coreference resolution.These models and methods are verified in the automatic construction of physics education ontology for middle school physics.