Iterative Statistical Language Model Generation for Use with an Agent-Oriented Natural Language Interface

摘要: We describe a method for developing statistical language model (SLM) with high keyword spotting accuracy natural interface (NLI). The NLI is based on the Adaptive Agent Oriented Software Architecture (AAOSA). Our experience shows that this provides rapid development of an SLM well suited to requirements agent-oriented NLI. Experiment results show comparatively low equal error rate 13.2% vocabulary 2400 keywords. This result robust free-form speech-based task completion rate.