The challenge of spoken language systems: Research directions for the nineties

摘要: A spoken language system combines speech recognition, natural processing and human interface technology. It functions by recognizing the person's words, interpreting sequence of words to obtain a meaning in terms application, providing an appropriate response back user. Potential applications systems range from simple tasks, such as retrieving information existing database (traffic reports, airline schedules), interactive problem solving tasks involving complex planning reasoning (travel planning, traffic routing), support for multilingual interactions. We examine eight key areas which basic research is needed produce systems: (1) robust recognition; (2) automatic training adaptation; (3) spontaneous speech; (4) dialogue models; (5) generation; (6) synthesis (7) systems; (8) multimodal systems. In each area, we identify challenges, infrastructure research, expected benefits. conclude reviewing need multidisciplinary development shared corpora related resources, computational far rapid communication among researchers. The successful this technology will increase accessibility computers wide users, facilitate multinational trade, create new specialties jobs rapidly expanding area. >