Web service technology allows access to advertised services despite of their location and implementation platform. However, considerable differences on structural, semantical and technical levels along with the growing number of available web services makes their discovery a significant challenge. Keyword-based matchmaking methods can help users to locate quickly the set of potentially useful services, but they are insufficient for automatic retrieval. On the other hand, the high cost of formal ontology-based methods alienates service designers from their use in practice. Several information retrieval approaches to assess the similarity of web services have been proposed. In this paper we proceed with such a study. In particular, we examine advantages of using Vector-Space Model, WordNet and semantic similarity metrics for this purpose. A matching algorithm relying on the above techniques is presented and an experimental study to choose the most effective approach is provided.