Abstract: A bullet-comment type recognition model is constructed based on semantic features, and then the impact of the bullet-comment type on online consumption behavior is explored on the basis of prediction and analysis of types of bullet-comments. The results show that: there is a “herd effect” in the influence of bullet-comments on the process of online consumption behavior. On the whole, the volume and emotion of bullet-comments influence the process of consumption behavior. The volume and emotion of content-related bullet-comments can stimulate users’ experience and trust, and further promote the deep purchase behavior through improvement of post-purchase experience. The volume and emotion of sentiment-related bullet-comments negatively affect individual experience. The increase in volume of communication-related bullet-comments has a positive effect on sharing behavior. The above conclusions enrich the research about instant comments and also have important practical value for the advertising design and the management of bullet-comments.

Key words: content-related bullet-comment, sentiment-related bullet-comment, communication-related bullet-comment, online consumption behavior