| 英文摘要 | The elements of news are the summarization of the time, the location, the participants, the main events, the procedure and the causes of the news fact, which are usually noted as 6Ws. They are powerful instruments found from long-term work and practice by journalists for people to grasp and describe the news fact. The analysis of elements is fundamental to the success of semantic understanding, and is critical to effective enhancements to various natural language processing technologies, such as information extraction, text summarization, question answering, machine translation, etc. And now both theoretical and technical researches on the analysis of the elements of modern Chinese news are pressing under the circumstances of the continual popularization of Chinese and the rapid spreading of news through the internet and other media. The purpose of research on the semantic analysis of news elements is to catch the elements depicting the fact from a news sentence and thereby to facilitate the understanding of the sentence. Based on the concept knowledge tree model, we have, in this thesis, made a comprehensive study of the important aspects of the elements of Chinese news, which includes the semantic characteristics, the semantic representation model, the organization of related knowledge, and the automatic semantic analysis algorithm. The main contributions of this thesis include following issues: 1. In this thesis we bring the concept of news elements (When, Where, Who, What, How and Why) into the domain of natural language understanding from mass communication, investigate its domain independent attribute, devote the first four elements to semantic parsing of sentences which describe news fact, and consequently we obtain a more general domain independent approach to describe semantic information compared to the semantic roles. 2. Concept knowledge tree model is an architecture based on concept for knowledge representation and organization. Besides the achievements already in existence, we give the formalism of this model, propose a mean state driven algorithm for compounding concepts with the help of elements analysis，and thus make the model much sounder. 3. We propose a semantic representation model of news elements in consideration of its syntactic and semantic characteristics under the semantic representation frame which belongs to the concept knowledge tree model and is based on concepts. We use the concept model and its internal structure to re... |
修改评论