CASIS: a context-aware speech interface system

摘要: In this paper, we propose a robust natural language interface called CASIS for controlling devices in an intelligent environment. is novel sense that it integrates physical context acquired from the sensors embedded environment with traditionally used to reduce system error rate and disambiguate deictic references elliptical inputs. The n-best result of speech recognizer re-ranked by score calculated using Bayesian network consisting information input utterance context. our prototype uses device states, brightness, speaker location, chair occupancy, direction action history as context, has been reduced 41% compared baseline does not leverage on information.