By exploiting existing processing tools, it is sometimes possible to perform complicated computer processing that could not be justified de novo. Using an existing parsing program and dictionaries as a starting point, a set of programs for building a bibliographic dictionary automatically from text was generated and then tested on a newswire text. In general, the error rate of output records is less than 10 per cent. Whether this quality of output is useful can only be judged in the context of a definite application.