Accessible technology for interactive systems: a new approach to spoken language research

In this paper. we argue for a paradigm shift in spoken language technology, from transcription tasks to interactive systems. The current paradigm evaluates speech recognition technology in terms of word recognition accuracy on large vocabulary transcription tasks? such as telephone conversations or media broadcasts. Systems are evaluated in international competitions, with strict rules for participation and well-defined evaluation metrics. Participation in these competitions is limited to a few elite laboratories that have the resources to develop and field systems. We propose a new, more productive and more accessible paradigm for spoken language research, in which research advances are evaluated in the context of interactive systems that allow people to perform useful tasks, such as accessing information from the World Wide Web, while driving a car. These systems are made available for daily use by ordinary citizens through telephone networks or placement in easily accessible kiosks in public institutions. It is argued [1,2,3] that this new paradigm, which focuses on the goal of universal access to information for all people, better serves the needs of the research community, as well as the welfare of our citizens. We discuss the challenges and rewards of an interactive system approach to spoken language research, and discuss our initial attempts to stimulate a paradigm shift and engage a large community of researchers through free distribution of the CSLU Toolkit. 1. SPOKEN LANGUAGE SYSTEMS Spoken language systems allow people to interact with machines using speech to accomplish useful tasks. The essence of a spoken language system is interaction-the dynamic interaction between a person and a machine using speech, and the interaction of the different language technologies within the system. At a minimum, a spoken language system integrates dialogue modeling, speech recognition and speech generation. It can also include natural language understanding, language identification, machine translation, speaker recognition, as well as other multimodal (e.g., handwriting, gesture recognition, speech reading) and multimedia (e.g., facial animation, video) capabilities. The success of a spoken language system depends upon the manner in which the component technologies interact to produce an effective dialogue that accomplishes the task at hand. An effective system produces prompts that elicit the set of desired responses from the user (and minimizes undesired responses), detects recognition errors and out of vocabulary utterances, engages in conversational repair when such errors occur, and responds in an appropriate wa,v when the dialogue breaks down. While performance of each component technology is important. the manner in which they interact is even more so. Speech recognition is but one essential component of an integrated system. To use an analogy, it is well understood that there is little gain in increasing the processor speed in a computer, when the processor is starved of data. In that case one should speed up the data access before increases in processor speed will be of benefit. Similarly, in spoken language systems, components other than recognition will at some point mask any improvements in recognition. The interactions among the modules of spoken language systems are usually highly complex and interdependent and can be studied and understood only by developing and evaluating working systems. Based on their experiences in developing a spoken language system for taking the U.S. census, Cole et al. [6] conclude: “Taken together. the results of this project showed that the most important component of a spoken dialogue system is the dialogue. A successful system gives instructions efficiently. establishes expectations for the user, asks questions that constrain the possible responses. and proceeds in a straightforward manner to complete the interview.” 2. THE NEED FOR SPOKEN LANGUAGE