商品评论是消费者针对某一个商品的主观议论。针对微博中商品的评论文本短小、结构多样等特征,在仅使用现有的微博级情感标注的条件下,提出了一种基于层叠条件随机场模型。以中文小句中枢说为理论基础,将商品评论的句子划分为若干小句,使用微博内小句序列的各种特征训练粗粒度的随机条件场情感分类模型,同时使用小句内汉字序列的各种特征来训练细粒度的随机条件场情感分类模型。实验结果表明,本文提出的方法优于传统的情感分类方法。 Product reviews are consumer subjective arguments about a product. Aiming at the short texts and diversified structure of the articles in Weibo, this paper proposes a stochastic field model based on cascade conditions, using only the existing Weibo-level emotion labeling. The Chinese clause center is taken as the theoretical foundation, the sentence of commodity comment is divided into several small clauses, and the coarse granularity random emotion field emotion classification model is trained by using the features of small sentence sequences in Weibo. At the same time, To train fine-grained stochastic conditional sentiment classification models. Experimental results show that the proposed method is superior to the traditional method of emotion classification.