Three existing applications that integrate a level of speech recognition, natural language processing and animation have been discussed. QuickSet integrated all three in a multi-agent architecture, however it was a multi-modal interface and utilised gesture as well as speech input.
CommandTalk falls closer to the scope of this paper in that it does not incorporate any input modes other than spoken natural language. It has an advanced NLP module allowing clarification dialogue to remove several classes of ambiguity.
Finally, AnimNL which uses typed rather than spoken natural language as input was reviewed. It focuses on the planning and evolving of understanding during the action of required tasks.
|
Table 2.5 shows a summary of the existing applications, and the technologies they integrate. As can be seen QuickSet has the largest feature set as it is a multimodal system connected to a battlefield simulator. CommandTalk integrates speech recognition, a high level of interactive natural language processing, and a similar battlefield simulator. AnimNL concentrates on the animation of a single character rather than many units involved in combat. It does not however use any form of speech recognition.
