Abstract
In this paper we investigate the use of probabilistic machine-learning techniques to extract latent factors from semantically enriched service descriptions. The latent factors provide a model to represent service descriptions of any type in vector form. With this conversion, heterogeneous service descriptions can be represented on the same homogeneous plane thus achieving interoperability between different service description technologies. Automated service discovery and ranking is achieved by extracting latent factors from queries and representing the queries in vector form. Vector algebra can then be used to match services to the query. This approach is scalable to large service repositories and provides an efficient mechanism for publishing new services after the system is deployed. © 2011 IEEE.