Detecting User Occupations on Microblogging Platforms: An Experimental Study

User occupation refers to the professional position of a user in real world. It is very helpful for a number of applications, e.g., personalized recommendation and targeted advertising. However, because of the risk of privacy leaks, many users do not provide their occupation information on microblogging platforms. This makes it hard to detect user occupations on microblogging platforms. In this paper, we conduct an experimental study on this issue. Particularly, we propose an experimental framework of detecting user occupations on microblogging platforms. We first implement a number of classification models and devise various sets of features for user occupation detection. Then, we propose to construct an occupation-oriented lexicon, which is collected by an iterative extension algorithm considering semantic similarity and importance between words. We combine the lexicon with the word embedding approach to detect user occupations. We conduct comprehensive experiments and present a set of experimental results. The results show that the lexicon-based word embedding method achieves higher accuracy compared with traditional feature-base classification models.