At present, question-answer (QA) sites have become one of the most important sources of information sharing. In order to ease search and categorization, QA sites create communities to discuss a specific topic or interest. As a consequence, a large number of communities have been created in the last few years. A lot of research has been conducted on community QA sites to address various problems including expert identification and tag recommendation. However, an important problem that has been neglected so far is to automatically route a question to the right community. In this paper, we propose a novel word-embedding based method to route a question to the right community. We use syntactic as well as semantic features to characterize a question and community. Although this approach of characterization performs well, it is highly computationally expensive. To deal with this problem, we use topic modeling, which effectively summarizes a community and reduces the computation time. Our experimental results reveal that usage of both syntactic and semantic features helps in question routing and leads to a better community prediction. We evaluate our methods on a well-known question answering system Stack Exchange and show the effectiveness of the proposed method. © 2019 Association for Computing Machinery.