Enhancing Access to Situational Vocabulary by Leveraging Geographic Context

Users of augmentative and alternative communication (AAC) aids could benefit from novel methods for accelerating access to contextually relevant vocabulary. This paper describes our initial efforts toward improving access to situational vocabulary through the use of geographic context to predict vocabulary. A corpus of spoken data produced by one able-bodied male speaker across various locations was collected and used to mine location-specific vocabulary clusters. We describe the data mining algorithm and illustrate how context-driven vocabulary organization and prediction can be integrated into an iconic communication system. Correlations between user and algorithmgenerated vocabulary patterns highlight discrepancies between actual versus perceived frequency statistics. Thus, device customization may be achieved through a hybrid humanalgorithmic approach to vocabulary selection. Endowing the communication aid with knowledge about the user’s location-specific vocabulary patterns may help to balance the burden of communication between the system and the user to yield substantial gains in accessing situationally appropriate vocabulary which may in turn accelerate communication rate.