ChIMP: Children Interacting with Machines Project

Spoken language adds naturalness and efficiency to human-machine interactions with both children and adults. The long-term goal of this inter-disciplinary research project is to enable spoken language interaction capability in immersive environments through better understanding of how humans converse with machines. Two crucial requirements for conversational humanmachine interfaces are robustness (i.e., minimal performance degradation under trying conditions) and adaptability (i.e., responding to the user in a customized manner). The near-term research focuses on two specific issues related to automatic speech recognition and understanding: (1) The challenge of recovering from system errors made while interacting with capable adult users, and (2) the challenge of adapting to the special characteristics of childusers within a developmental framework. The immediate goal of the project is to analyze two previously collected corpora of spoken human-machine communication along a number of linguistic dimensions, and to use our findings to create realistic user models for spoken language interactions.

NSF Report
Poster
More Info
Laboratory