In this IRES project supported by the Office of International Science and Engineering, U.S. graduate and undergraduate students under the direction of Professor Avelino Gonzalez at the University of Central Florida will participate in a collaborative research effort in conversational avatars with Dr. Patrick Brezillon at Universite de Pierre et Marie Curie in Paris, France. The U.S. students will explore the possibility of using contextual graphs (CxGs) to model and develop a tool for identifying the context of conversations and improving dialog management.
Conversational avatars are artifacts, such as computer programs, robots, or animations, which behave and communicate like specific people. Such avatars have potential applications across many fields, including health, business, and defense. One example is in the area of health maintenance, where improvement in the naturalness of avatars could improve their acceptance by senior citizens who need human-like coaching in therapy or medicine management. In the course of working on their research projects, the students will learn both how to speak French as well as how to interact in French culture. They will thus gain an invaluable research experience in an international setting, helping contribute to the development of a new generation of a skilled technical workforce with a global perspective.
It is a well-known fact that modern scientific and engineering research knows no national boundaries. This has been true for a long time, as scientists from different countries have collaborated from as far back as the 18th century. However, the recent advances in electronic communication have enabled a level of collaboration not previusly possible. Not only are advances in research made throughout the world, but many times, teams of scientists and engineers from different nations will collaborate on research that could not be done by any one of the team members alone. For students interested in pursuing a career in STEM research, understanding this concept is not only important but also critical. Yet, there are many aspects to international research that cannot be learned by simply reading books or listening to lectures – they must be experienced. It was this very experience that we sought to provide our students as part of this grant in the Department of Electrical Engineering and Computer Science at the University of Central Florida. Eleven students from theUniversity of Central Florida paticipated in this project. Four were undergraduates in computer science; three were undergraduates in computer engineering, one was an undergraduate in electrical engineering, one was an undergraduate in psychology and two were PhD students in computer engineering. All students participated in short term visits to our partner institution, the Universite de Pierre et Marie Curie (UPMC) in Paris, as well as other sites of significance in our area of research. Four of them participated in long-term visits to UPMC. To a person, these students profitted enormously from the experiences and the collaboration with researchers at UPMC. In addition, the group made significant intellectual advances in the state of the art in conversational avatars. In collaboration with the University of Illinois - Chicago, we had previously (under a different NSF grant) developed lifelike avatars that resembled actual human beings, and with whom one could converse in natural speech in a question-and-answer context about some specific topic. We leveraged this technology with the contextual graphs conceived by our collaborator at UPMC, Dr. Patrick Brezillon, to improve the dialog capability of our Lifelike avatar. The most sinificant was the creation of a dialog management system that incorporates contextual graphs (CxGs) to further refine the granularity of the knowledge. When combining CxGs with our prior dialog management system based on Context-based Reasoning (CxBR), this allowed the avatar to answer very specific questions. With the former system, it could only answer general questions very generally. Furthermore, the group incorporated the SART system - an intelligent assistant for incident management in the operation of the Paris metro system - as part of the question-and-answer avatar. A tool was created to facilitate the creation of the CxGs necessary for our dialog managemrnt system. While one such tool had already been developed by our partners at UPMC, it did not satisfy our needs for the enhancements that we made to the CxG paradigm to make it synergistic with our dialog management system. Thus, the need for us to develop our own tool. Lastly, the Lifelike avatar was made able to "speak" and "understand" French. This made the reearch bi-cultural in nature. Other research projects of significantly smaller scope were undertaken by the students as part of this project. The major research objectives of the project were mainly undertaken by the students who participated in the long term visits to UPMC. However, each student had to have at least one research project in order to enjoy a hands-on experience as well as receive direct mentoring by the project PI, Co-PI and/or our collaborator at UPMC, Dr. Brezillon. Having and making progress in a personal research project was a condition for traveling to UPMC and/or the other related sites. The collaboration has directly led to the UCF PI (Gonzalez) and the UPMC collaborator (Brezillon) to co-edit a contributed volume titled "Context in Computing". While not directly related to the tasks completed under this grant, it is something that grew out of this collaboration, and a clear sign of a closer relationship with UPMC that resulted from this grant. Two of the students who participated in this project (Hung and J. Parker) have had manuscripts accepted for inclusion into this volume.
