Sentences Extraction from Digital Publication for Domain-Specific Knowledge Service

Digital publication resources contain a lot of useful and authoritative information which is normally organized in small sections such as paragraphs, book sections or chapters. It is important to use the information from digital publication resources for knowledge service. In this paper, concepts in a domain are obtained from encyclopedia. Sections are extracted from e-books and then indexed for searching. The related sections for the important concepts are then found by using full text search technique. SVM is used to classify the related sections and the semantic information is computed for the concept. The sentences are then extracted by dynamically extending the adjacent sentences into sentence group. With the method, the sentences extracted are continuous and the length of the sentences would approximate to a specified length statistically. The method is effective for domain-specific knowledge service.