Recently, spatial keyword queries (SKQ) have become a hot topic in database field. However, Most of the existing SKQ methods are limited in Euclidean space or assume that objects (and queries) are static. This paper addresses the issue of processing continuous top-k spatial keyword queries over moving objects (CMTkSK) in road networks. To efficiently index moving geo-textual objects in road networks, a novel index structure called TPRgt-tree is proposed. Based on the index, an efficient CMTkSK query processing method which includes three main phases, namely generating initial result set phase, pruning phase, and continuous monitoring phase, is proposed. The proposed method can deal with the situation where the query client and geo-textual objects move continuously in the road network. By finding the result change time points, the method can continuously monitor CMTkSK queries and keep the query result set up-to-date with a small price. Finally, experiment results show that the proposed method is much more efficient and precise than its competitor.