A method is provided of accessing a database comprising a plurality of entries, each entry comprising a main entry containing a headword. At least one of the entries comprises a first subsidiary entry (2) containing a plurality of sets of words, each of which is capable of being associated with the headword. The method comprises the steps of: entering a section of text which comprises a plurality of words; locating the entry whose main entry contains the headword corresponding to a selected word of the section of text; and, where the located entry has a first subsidiary entry containing a plurality of sets of words, selecting from the first subsidiary entry a set of words of which at least one word is present in at least part of the section of text.