基于LBS的中文自动文摘技术研究  

Research on Chinese Text Automatic Abstracting Technology Based on LBS

作　　者：张清军[1] 朱才连[1] 

机构地区：[1]中国科学院测量与地球物理研究所,湖北武汉430077

出　　处：《四川大学学报（工程科学版）》2004年第4期99-102,共4页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家自然科学基金资助项目(40274058)

摘　　要：LBS的终端用户通过各种无线手持设备访问因特网,获取与位置有关的资讯,但由于这些设备显示屏较小,再加上无线通讯网带宽不足,无法浏览整个网页,采用文本摘要来浓缩整个网页将是LBS中重要技术之一。提出了一种基于文本结构分析的文摘方法,首先通过向量空间模型来计算段落和全文的相似度,按照给定的阈值选定主题段落;然后计算主题段中各个句子与相应主题段的相似度,按照相似度由高到低选取主题句,组成粗的文摘。实际开发的系统原型验证了此方法的有效性。The LBS users retrieve location-relevant information from the World-Wide Web using small handheld devices. Because of small screen and narrow bandwidth, it is difficult for these devices to browse the entire Web page. So it is an important technology in LBS that automatically abstracting the entire Web page. In this paper, a new approach of automatic abstracting based on text structure is provided. Firstly, the similarity values between paragraphs and the full text are calculated by making use of Vector Space Model and extract the topic paragraphs according to the threshold value. Secondly, the similarity values between each topic paragraph and each sentence in the paragraph are determined. The topic sentences which constitute the abstract are extracted from the topic paragraphs according to the similarity values from high to low. A prototype is implemented and experimental results show that this new method is effective.

关 键 词：LBS 向量空间模型 自动文摘 

分 类 号：P209[天文地球—测绘科学与技术] TP391.1[自动化与计算机技术—计算机应用技术]