In the study reported in this paper, we apply a family of Information Retrieval (IR) models to overcome the problem of retrieving services, whose descriptions match users' queries given a free text style. This family is composed by four models which have not been applied in prior research on IR-based service discovery. Among these models, two of them are based on matrix factorisation models applied to Latent Semantic Indexing (LSI). Moreover, we propose other two models based on query expansion. The first one expands the query with terms retrieved from the lexicon called WordNet, taking into account parts of speech of the query. The second model also expands the query, but with terms extracted from an automatically generated co-occurrence thesaurus. The results of the experiments suggest that the last model outperforms the other models applied in this work besides those most recent and prominent in the state-of-the-art on IR-based service discovery.