FoG: Synthesizing forecast text directly from weather maps
Eli Goldberg · 2002
FoG (Forecast Generator) is a program which has been developed using natural language processing (NLP) techniques to model different varieties of weather forecast text. Text forecast products in both French and English are produced from a sequence of weather depiction charts. The goal is to relieve weather forecasters of the chore of manually composing routine text forecast messages. At the same time, FoG eliminates the translation of text products, and facilitates the production of more timely and more consistent forecast products. By calling upon NLP methods, FoG is able to successfully overcome the problems of software maintenance and flexibility which have always frustrated earlier computer worded forecast systems. There are three principle components to FoG: a conceptual component which extracts the required meteorological information from a sequence of weather maps, a text planner, and a realization component. The conceptual component is an expert system while the other two components rely on NLP techniques. FoG was written in Quintus PROLOG with object-oriented extensions.>