Recent advances in language processing and speech recognition open up a large opportunity for video game companies to embrace voice interaction as an intuitive feature and appealing game mechanics. However, speech-based systems still remain liable to recognition errors. These add a layer of challenge on top of the game's existing obstacles, preventing players from reaching their goals and thus often resulting in player frustration. This work investigates a novel method called context-aware speech recognition, where the game environment and actions are used as supplementary information to enhance recognition in a speech-based game. In a between-subject user study (N=40), we compared our proposed method with a standard method in which recognition is based only on the voice input without taking context into account. Our results indicate that our proposed method could improve the player experience and the usability of the speech system.
https://doi.org/10.1145/3613904.3642426
The ACM CHI Conference on Human Factors in Computing Systems (https://chi2024.acm.org/)