A system and method for multi-language phrase identification within spoken interaction audio capable of adjusting for regional pronunciation (accents), cadence differences, and homologs. In this system, a spoken interaction audio data store supplies spoken audio data such as contact center call recordings to be analyzed for a specific phrase or set of phrases. Phrases are entered as natural language text and converted to the phonemes representative of the phrase audio using the invention's language packs and stored in a data store. Spoken interaction and phrase audio are converted to a digital format allowing comparison using multiple characteristics. Phrase matches are stored for subsequent post analysis display and analytics generation.