Abstract: A reference table, which may not be stored, has columns associated with data attributes and rows containing related words assigned to those attributes in a collection of data. The stored data include thesauruses associated with the attributes, and reference table row identifier lists respectively associated with thesaurus entries. Each thesaurus is defined with reference to a partition into subsets of the words which can be assigned to the associated attribute, and has a respective entry for each subset including an assigned word, the row identifier list associated with this entry including any identifier allocated to a row of the reference table having a word of the subset assigned to the associated attribute. A matching reference table row identifier list is determined from the data representing the row identifier lists associated with thesaurus entries relevant to the query.