| Today, news information has become one of the most important types of information. In real life, people often want to get news theme information timely, and avoid to browse a lot of news reports, they only interested in their concerned subject. The news information management and retrieval model did not take full advantage of the characteristics of news information, and take news information as plain text content, can not meet the needs of the public. With the arrival of the era of new media, people put forward new requirements for quick browsing news reports.Topic detection technology can automatically detect the latest news from the news stream, and news reports organized in a timely manner according to the news topics, then can get the special news reports. Therefore, topic detection technology in the application will be able to effectively manage and organize news information, meet the special needs of people. The information industry in ethnic minority area will rapidly develop in the future, there are a lot of related news in the network, to develop a set of intelligent system to handle these information are very useful. Moreover, further study of topic detection technology of Uyghur can also provide important reference for related technology research of other ethnic languages.This paper analyzes development present situation of the topic detection and tracking technology at home and abroad, and designs the topic detection system for Uyghur news texts. Among them, the news feature extraction for news elements, then using the adding window strategy and dynamic incremental clustering algorithm, and improved the window strategy; When comparing the improve suffix tree clustering with the traditional suffix tree clustering, the accurate and overall rate of the improved algorithm are improved; For the affinity propagation algorithm, we take some corresponding modification for matrix and the parameters, in order to adapt to the news topic detection, on the basis of the stability of the algorithm it can improve the accuracy of the algorithm; Finally, the paper takes the single-pass algorithm as a benchmark, compares with several main topic detection algorithm, analyzes the advantages and disadvantages of these algorithms and give experimental verification.According to the characteristics of Uyghur news, the paper proposes a news detection model based on news topics. The model uses the topic detection technology to detect news topics from the news stream, and analyzes related topic elements like named entities, the clustering algorithm can adapt the features of news information, and get the better detection results and personalized service. On the basis of news topics, a news topic detection prototype system is designed and realized.The paper also has carried on the related experiments, and get the experimental results. |
PDF Full Text Request