一种面向e-Science环境的多领域Web文本特征抽取模型.PDFVIP

2011 1 1 Journalof Chinese Computer System s Vol32 No. 1 2011 eScienceW eb 1, 2 1 1 1 翁 彧 , 胡长军 , 席 强, 张学春 1 (, 100083) 2 (, 100081) Emai:l mr. wengyu@ gmai.l com : 传统领域信息抽取方法多依赖领域词典实现文本特征的发现, 既不便于实验复现, 也不易于 在多领域环境中移植与 推广, 严重制约了模型的应用范围.针对上述不足, 提出 一种适用于 eScience环境的多领域 W eb文本特征抽取模型(简称 e WTDE) . 该模型将无词典分词技术引入多领域文本特征发现过程,摆脱了对于领域词典的依赖; 借助对领域主题及 具体事件 中共性与 个性特征的抽取与分类, 模型动态追踪领域事件发生及 发展变化,并最终形成多 个区域性数据中心; 通过对各数据 中心中领域知识的协同调度, 有力提高了领域信息在全局范围内的利用效率.验证实验中分别对多领域特征抽取主题特征动 态追踪以及领域知识协同调度予以有效性验证, 并进一步证明了模型的实用效果. : eScience; ;; W eb : TP311 : A : 2011) M ultidomainW eb Text Feature ExtractionM odel for eScience Environm ent 1, 2 1 1 1 WENG Y u , HU Changjun , X I Q iang , ZHANG X uechun 1 ( School of Information and Eng ineering, U niversity of Science and Technology eijing, eijing 100083, China ) 2 (College of Information Engineering,M inzu University of China, eijing 100081, China ) Abstract: The traditional inform ation extractionm ethods based on specific dom ain usually depend on the dom ain dictionaries to dis cover the text feature. It is inconvenient for reproducing and difficult to transplant in multidom ain environment. The application scope is lmi ited seriously. O riented to the deficiencies above, a multidom ain w eb text feature extractionmodel for eScience is pro posed ( nam ed eW TDE). Thismodel adopts the Chinese splitw ords techno logy w ithout dictionary into the processof m ultidomain text feature discovery and avo ids the dependency of dom ain dictionaries effectively. W ith the help of classification of common and in dividual features, them odel tracks the generation and the development trend of domain events dynam ically, and form s a couple of lo cal data centers eventually. Through cooperative scheduling the domain know