| With the wide application of the personal computer and network,and the development of OCR(optical character recognition)technology,it becomes very convenient to transform paper documents into digital documents,which facilitates the sharp increase of digital books on the Internet.It is convenient for users to get knowledge from plentiful digital books,but meanwhile it leads to the problem of information overload.Users would get thousands of related books when they search for a specific topic.However,different books have different organizational styles due to different authors.Therefore it is a very meaningful task of how to do incorporation and mine a clear learning venation among massive digital books.In order to improve efficiency of knowledge learning,we design the Knowledge Venation service system.We carry out knowledge mining on thousands of books related to the given topic,and summarize several learning paths with rich knowledge,good fluency and high coverage.The knowledge venation is composed of these learning paths,and is visualized with a metro map to help users to learn knowledge efficiently.Our contributions are as follows:(1)We implement a universal digital books structure analysis and processing scheme.TOC(table of contents)and paragraphs of OCR documents are recognized and extracted by an algorithm with features:document layout,visual and content.Finally,we get structured digital books.(2)We propose an unsupervised algorithm based on weighted word embedding to solve the problem of short texts matching.We can obtain learning objects by clustering the semantical similar chapters via an unsupervised clustering method.(3)We propose a learning path selection algorithm to select a collection of high informative and fluent but low redundant learning paths from the learning graph.The knowledge venation is constructed with these learning paths.(4)We use metro map to visualize the knowledge venation,and propose a Knowledge Venation service system. |
PDF Full Text Request