Abstract: The subject matter of this specification can be implemented in, among other things, a computer-implemented method for identifying and storing stem variants of search query terms. The method includes receiving a first query and a user selection of a first search result from the first query. The method further includes receiving a second query and a user selection of a second search result from the second query. The method further includes determining that the first search result and the second search result have the same category. The method further includes identifying a first term in the first query and a second term in the second query as stem variants of one another by determining that the first term is within a threshold edit distance of the second term. The method further includes storing the first term and the second term in a stem-variant repository.