Along with the growing popularity of cloud computing technology, the amount of available cloud services and their usage frequency are increasing. In order to provide a mechanism for the efficient enforcement of service-relevant operations in cloud environment, such as service discovery, service provision, and service management, a completed and precise service specification model is highly required. In this paper, we conducted a survey on existing service description languages applied in three different domains-general services, Web/SOA services, and cloud services. We discussed and compared the past literature from seven major aspects, which are: (1) domain, (2) coverage, (3) purpose, (4) representation, (5) semantic expressivity, (6) intended users, and (7) features. Additionally, two core dimensions-semantic expressivity and coverage-are employed to categorize and analyse the key service description languages by using Magic Quadrant methodology. These two dimensions are regarded as the most essential factors for the evaluation of a service description model. Based on this analysis, we concluded that Unified Service Description Language (USDL) is the language with the widest coverage from business, technical and operational aspects, while OWLS is the one that has the highest semantic expressivity. At last, critical research issues on cloud service description languages are identified and analysed. The solution of these issues requires more research efforts on the standardization of cloud service specification, which will eventually enhance the development of cloud industry.