Traditional dialogue systems are designed to provide access to a relatively structured and static back-end database (such as airline reservation information), where users have well-defined, task-oriented goals. Compared to such systems, building a spoken dialogue system to a chat environment is novel in many ways, and raises interesting challenges for dialogue system design. To explore these issues, we have designed and implemented CobotDS (Cobot Dialogue System). CobotDS extends ongoing work on Cobot, a software agent that resides in a well-known internet chat server called LambdaMOO. Founded in 1990, LambdaMOO is frequented by hundreds of users who converse with each other using both natural language text and verbs for expressing (in text) common real-world gestures (such as laughing, hugging, nodding and many others). Cobot is one of the most popular LambdaMOO residents, and both chats with human users, and provides them with ``social statistics'' summarizing their usage of verbs and interactions with other users (such as who they interact with, who are the most ``popular'' users, and so on).
CobotDS provides LambdaMOO users with real-time spoken telephony access to Cobot, and is an experiment in providing a rich social connection between a telephone user and the text-based LambdaMOO users. To support conversation, CobotDS passes messages and verbs from the phone user to LambdaMOO users (via automatic speech recognition, or ASR), and from LambdaMOO to the phone user (via text-to-speech, or TTS). CobotDS also provides ``listening'' (allowing phone users to hear a description of all LambdaMOO activity), chat summarization and filtering, personalized grammars, and many other features. We hope that CobotDS will provide an alternate means of access to LambdaMOO - either out of necessity (such as when a user is unable to access the internet), out of a desire to use a different input modality (speech instead of typing), or as an entertaining accompaniment to logging on directly. We also believe that our experiences may hold lessons for future attempts to provide spoken access to text systems (such as instant messaging), and more generally for multimodal systems.