Recent work at the Cambridge Language Research Unit has been concerned with the development of automatic classification procedures for information retrieval. This has taken the form of research into methods of classification of keywords extracted from documents, with a view to using the classes found for coordinate indexing of technical material. We cannot claim to have solved this problem because the methods we have been able to develop so far cannot be applied on a sufficiently large scale. We have, however, made enough progress to make us feel that this is a fruitful line of research.