Users currently expect to query data in a Google-like style, by simply typing some terms, called keywords, and leaving it to the system to retrieve the data that best match the set of keywords. The scenario is quite different in database management systems, where users need to know sophisticated query languages to retrieve data, and in database applications, where the user interfaces are designed as a stack of pages with numerous boxes that the user must fill with his search parameters. This thesis describes an algorithm and a framework designed to support keywordbased queries for datasets with schema, specifically RDF datasets and relational databases. The algorithm first translates a keyword-based query into an abstract query, and then compiles the abstract query into a SPARQL or a SQL query such that each result of the SPARQL (resp. SQL) query is an answer for the keywordbased query. It explores the schema to avoid user intervention during the translation process and offers a feedback mechanism to generate new answers. The thesis concludes with experiments over the Mondial, IMDb, and Musicbrainz databases. The proposed translation algorithm achieves satisfactory results and good performance for the benchmarks. The experiments also compare the RDF and the relational alternatives.