Model-Driven, Assertion-Based Generation of Multimedia Weather Information

Abstract

As gridded forecasts are becoming a common source of weather data, efforts are being made to automatically produce displays for various categories of decision makers such as emergency managers and road traffic officers. This paper proposes a new systematic approach to the generation of multimedia weather reports tailored to the specific needs of the user. The process consists of two steps: extraction of assertions from gridded datasets, and their presentation in the form of maps, charts, tables, and natural language text. The assertions organized by discourse relations relate weather elements to regions and periods of interest to the user. The user can explore the situation in a hypermedia mode by descending from summary surveillance displays to more detailed exploratory ones. This approach provides a structure for a class of applications dealing with the dissemination of weather information and supports them with general methods of assertion extraction and media generation. The application-specific knowledge is localized in five orthogonal and easily maintained models: territory, time, parameters, phenomena, and events. A prototype system has demonstrated the variability of the concepts.