Biologists exchange findings through an ever increasing number of on-line journal articles. Thus, tools that help to discover information in the extensive biological literature are desirable. This has led to increasing interest in text mining techniques. Protein interactions are of great importance in many biological processes, and therefore of great interest to biologists. The majority of the existent text mining approaches ~oncentrate on the extraction of positive protein interactions from the literature. However, it is also useful to know if a given set of proteins reportedly does not interact. Such additional information may, for instance, help biologists to avoid redundant experiments.