相对于传统的产品领域意见挖掘研究,文章对中文通用领域的意见挖掘各部分内容进行了尝试性研究。利用基于多种语言特征和候选评价对象的条件随机场模型进行观点表达抽取,对有窗口限制的最近邻方法进行改进,提出一种评价对象—观点表达对的匹配算法,其对评价对象抽取效果也进行了进一步的修正。
Relative to the traditional research on opinion mining in the product area,this paper makes an experimental study of various part of content of opinion mining in universal Chinese area.The Conditional Random Fields(CRFs) model based on multilingual characteristics and candidate evaluation object is used to extract the opinion expressions.The paper improves the window-constrained nearest neighboring algorithm,and proposes a matching algorithm for the pairs of evaluation object and opinion expressions.The paper further revises the extraction effect of evaluation object.