The Research And Application Of The Domain Specific Concept Extraction And Prerequis Ite Relat Ion Mining In Books

With the great changes in science and technology,the information technology has developed quickly.The massive books provide a convenient way of learning knowledge for human beings.Books contain a lot of authoritative and valuable information,so they have become one of the core methods to get new knowledge as well as enhance professional ability.However,there are more and more books and the quality of books can't be guaranteed.People are facing the problem of information overload.How to get the value from massive information for individuals is becoming an urgent need.In books,domain-specific concepts and their dependency relations are of great significance.Concepts can be used to summarize the content of books,so we can just learn the concepts to quickly understand the main topics of a book.We can also constitute a knowledge space by utilizing the dependency relations among concepts in books.Through the dependency relations among concepts,we can understand the knowledge better.Moreover,the dependency relations are also useful for curriculum recommendation,learning route planning and so on.In the dissertation,we model the dependency relations as the prerequisite relations.Therefore,it would be valuable to extract the domain-specific concepts and mine the prerequisite relations among concepts in the massive books.At first,we extract the domain-specific concepts in books automatically.Then,we mine the prerequisite relations among these concepts.Finally,We design and implement a knowledge space service system based on the domain-specific concepts and their prerequisite relations.Our main contributions are as follows:(1)We propose a method of extracting domain-specific concepts in books.At First,we extract high quality phrases as the concept candidates from massive books by utilizing keyword extraction algorithms.Then,we propose a graph-based ranking approach to re-rank the concept candidates to obtain the domain-specific concepts.(2)We propose an unsupervised method of mining prerequisite relations among concepts in books.Firstly,we extract the features for concept pairs and use the random forest algorithm as the classifier to establish a prerequisite relation classification model.Then,we establish an unsupervised prerequisite relation discovery model.Finally,the classification model and the discovery model will be executed iteratively until convergence.(3)We design and implement a knowledge space service system.We visualize the domain-specific concepts and their prerequisite relations,and also provide concept retrieve and path retrieve services.