Deep Comprehension, Automatic translation and Generation of Weather Reports (Weathra)

Weather forecasts were early noted to be a domain where automatic translation was possible (Kittredge, 1973). Everybody in the field knows that there is a computer in Montreal translating forecasts routinely be- tween French and English (METEO). The weather domain has proven to be a fruitful domain for further research as witnessed e.g. by the system for generating marine forecasts presented by Kittredge et al (1986), by the work by Goldberg et al (1988), by the system gene- rating public weather reports in Bulgarian reported on by Mitkov (1991) and the system translating Finnish marine forecasts into Swedish by Bl~tberg (1988). The Swedish Weathra system to be presented in this paper explores the language and semantics of weather forecasts further and it aims at deep comprehension of forecasts. Beside grammatical re- presentations, Weathra uses repre- sentations of the meteorological raw facts and secondary facts, e.g. the fact that it will probably rain at a place where there is a low pressure area. It uses a representation of meteorological objects with their properties as frames in a data base and graphic representation with tile standard meteorological icons on a map, e.g. icons for sun, cloudy, rain, snow, thunderstorm, westerly winds, L(ow) and H(igh) pressure, temperatures, e.g. 10-15. Weathra also features a dynamic discourse representation including the discourse objects which may be referred to by the words and anaphora in the text (cf Karttunen, 1976, Johnson & Kay, 1990). The discourse objects are regarded as instances of the (proto)types or (concepts), which are also available as frames in a database. The formal grammar, morpho- logy and lexicon of Weathra are based on experience from the machine translation system Swetra (Sigurd & Gawronska, 1988), which is also written in Prolog (LPA MacProlog). The Weathra system can understand weather forecasts in a fairly deep sense, depict its comprehension in a map, answer questions about the main contents and consequences, translate English forecasts into Swedish ones and vice versa, and generate various forecast texts in English or Swedish