A computing device receives a natural language query. The computing device determines, based on the natural language query, a set of search criteria. The computing device determines, from a plurality of data sets, two or more data sets that are relevant to the natural language query based at least in part on matches between the set of search criteria and attributes of the plurality of data sets. The computing device joins the two or more data sets, including joining one or more rows and one or more columns of the two or more data sets to produce a joint data set. The computing device outputs a representation of the joint data set.