We use data mining techniques for finding 82 previously unreported common proper motion pairs from the PPM-Extended catalogue. Special-purpose software automating the different phases of the process has been developed. The software simplifies the detection of the new pairs by integrating a set of basic operations over catalogues. The operations can be combined by the user in scripts representing different filtering criteria. This procedure facilitates testing the software and employing the same scripts for different projects.