STAMP: a suite of tools for analyzing multimodal system processing

In this paper we describe a new automated suite of tools for capturing and analyzing data on multimodal systems called STAMP. STAMP is designed to support research and development efforts for advancing next-generation multimodal systems. STAMP permits researchers to analyze multimodal system performance by: (1) recording data on users’ multimodal input and the system’s responding, (2) supporting flexible replay of these multimodal commands, along with n-best recognition lists for the individual modalities and their combined multimodal interpretation, and (3) supporting automated analysis using different metrics of multimodal system performance. This collection of tools currently is being used to conduct basic research on the characteristics of multimodal systems, and also to iterate different aspects of the Quickset multimodal architecture.