Although end-users have available a lot of on-line services to be con- sumed individually, it is their composed usage what has the potential to create new value-added services for end-users. In this sense, many efforts have been done to allow end-users to compose the services that they need by themselves. However, most of these solutions present two main problems: (1) they provide little support to help end-users to browse interminable lists of services, and (2) they present the blank piece of paper problem, which appears when end-users have to face an empty canvas to define a composition without any help to find the services that better fit their needs. In this paper, we present a solution to im- prove these problems by using natural language processing techniques in order to search and select the services end-users need to accomplish a specific goal. This solution has been implemented in the context of the EUCalipTool platform.