Vestec’s Call Steering system is designed for natural language menu navigation in call center environments. Compared to traditional natural language approaches, it increases call routing accuracy by 10-20% and reduces grammar development time and costs by 30-50%. It is ASR, language, and domain independent and can be used to convert existing DTMF menus to their natural speech equivalent as well as create new speak freely menus with any number of routing destinations.
The Call Steering system comes in two models: Knowledge Engine (KE) and Data Engine (DE). KE is designed for small number of routing destinations, does not require any training data, and generates a finite-state (FSG) grammar. DE is designed for large number of routing destinations, requires training data in the form of tagged transcriptions, and generates a statistical language model (SLM) grammar. Both systems utilize a unique garbage model to improve recognizer grammar and a powerful NLU engine to improve semantic interpretation. A comprehensive grammar tuning system is also provided.