Extracting object of comment is an important part of emotional analysis. In view of the irregularity of language and the characteristics of network in Chinese online comment, this paper presents a method of extracting objects based on the syntactic analysis and conditional random field. It analyzes experimentally the effect of different templates and different combinations of features on the F value. In the implementation of the system, this paper uses Harbin Institute of Technology language platform open interface and CRFs open source tools to train and test on comment data sets. Finally, the F values of the two types of data sets have reached 82.98% and 83.50% respectively.