【摘要】 信息抽取模式的自动获取不仅能极大地减少所需的工作量,而且对于信息抽取系统的可扩充性和可移植性都至关重要。该文提出的采用开放语料库的跨领域模式自动获取方法是采用开放的训练语料库来自动生成领域无关的抽取模式,借助模式和实例之间的二元性,不断获取新的语料对系统进行训练,从而克服以往训练语料库规模较小的缺点。该方法尽可能减少了领域相关性,领域移植的代价很小。由于采用开放语料,该文提出了一种通过机器学习使用统计模型来自动选取用于模式生成和评价的正反实例的方法。根据在投资领域的测试结果,该方法获得了较好的效果,模式抽取的平均查准率达到92%,平均召回率达到42.4%。更多还原

【Abstract】 Acquisition of Information Extraction (IE) patterns not only reduces t he manual effort, but also increases the adaptation of IE systems. In this paper , we present an approach to acquire domain portable patterns automatically from dynamic training corpus. The basic idea of the approach is based on duality betw een patterns and instances. This approach is domain portable, it can be used in different domain conveniently. Because of open training corpus, we use machine l earning to identify some positive and negative instances based on the statistica l model. It has been proved to be reasonable in investment domain. The average p recision for pattern acquisition is 92%, the average recall is 42.4%.更多还原