The vision of a worldwide computing network of services that Service Oriented Computing paradigm and its most popular materialization, namely Web Service technologies, promote is a victim of its own success. As the number of publicly available services grows, discovering proper services is similar to finding a needle in a haystack. Different approaches aim at making discovery more accurate and even automatic. However they impose heavy modifications over current Web Service infrastructures and require developers to invest much effort into publishing and describing their services and needs. So far, the acceptance of this paradigm is mainly limited by the high costs associated with connecting service providers and consumers. This paper presents WSQBE+, an approach to make Web Service publication and discovery easier. WSQBE+ combines open standards and popular best practices for using external Web services with textmining and machine learning techniques. We describe our approach and empirically evaluate it in terms of retrieval effectiveness and processing time, by using a data-set of 391 public services.