Enemy items are two test items that should not be presented to a candidate on the same test. Identifying enemies is essential for personnel assessment, as they weaken the measurement precision and validity of a test. In this research, we examined the effectiveness of lexical and semantic natural language processing techniques for identifying enemy items. Within each technique, three item conceptualizations were compared: Stem, stem and key, and stem and all response options. The results suggest that lexical and semantic techniques are not interchangeable, and the semantic technique with item stem and key functioned most like subject matter experts. Details on how both techniques were implemented, and a discussion of the implications, limitations, and future directions of this research are provided.