International surveys lead to multilingual open-ended responses. By grouping the answers for the same categories from different countries, sets of comparable category-documents are obtained. The methodology here presented, called multiple factor analysis of contingency tables (MFACT), provides a representation of all of the documents in a common reference space, on the one hand, and of all of the words in a common reference space, on the other hand; these both spaces are related by transition formulae. This methodology allows a direct generalized canonical analysis approach to analyzing multiple contingency tables as well as for keeping correspondence analysis-like features. It works from the raw data, without any previous translation. An example, extracted from a large international survey in four countries, illustrates the potential of this approach.