Adaptive Dialog and Runtime Engine
Modern trends in care delivery show the need for close monitoring: therapeutic decisions should be taken as soon as possible basing on effects observed or reported by patients, and the monitoring should not incur in higher costs. The low cost and high availability of automated dialog systems make such systems promising candidates for a solution to the dilemma.
AdaRTE (Adaptable Dialogue Architecture and Runtime Engine) is a novel architecture for dialogue development. It provides dialog developers with a centralized, high level dialog representation; the platform handles the the actual dialog interaction, adding advanced features such as access to external resources, state-of-the-art dialog strategies, e.g. automatic adaptation to the user's experience.
The system is in operation on several real-world applications for management of chronic patients. Automated dialog systems have been deployed for hypertension, tele-dialysis, diabetes management, and oral anticoagulant therapy, and others. They are designed and tested in collaboration with several clinical partners.
AdaRTE combines the platform independence and the ease of web development granted by VoiceXML with the control flow and dynamic behavior provided by object-oriented programming languages.
AdaRTE addresses the following objectives:
- Streamline and automate dialog development cycle, allowing for rapid prototyping.
- Support frequent data transfer between dialogue system and other external resources such as databases at runtime.
- Avoid proprietary ASR engine solutions (voice platform independence)
- Facilitate the design of mixed initiative dialogues and incorporation of dialog development best practices
- Provide a platform for alternative output media incuding non-verbal interfaces, in addition to speech and touch-tone input: multimodality.
The AdaRTE framework builds on the experience gathered developing and deploying real-world adaptive dialogs and multiaccess systems for the care of chronic diseases.
AdaRTE is an active R&D project at the Laboratory for Biomedical Informatics (Università degli Studi di Pavia).
In collaboration with Amuser, Istituto Neurologico Mondino, Ospedale S. Matteo, and others.
Labmedinfo is looking for partnerships for:
- design of new dialogues to support clinical routine tasks
- reselling dialog systems for the home-care
- licensing the dialog interpretation system
Realizzazione di un plugin per lo sviluppo di interfacce multimodali (read announcement ).
An AdaRTE Graphical Interface, describing graphically the dialogue flow as a finite state network (5 years' dissertation).
Integration of multimodal capabilities as animated avatars (5 years).
Design, layout, implementation and testing of dialogues and grammars for various patient monitoring tasks (3-5 years).
Design and implementation of software Quality Assurance (QA) test cases.
- Mixed initiative dialogues, adaptivity, dialog best practices
- Reusable dialog components
- W3C standards: VoiceXML, SGRF, SISR formats