Generating simple weather reports in Maltese

Abstract

This dissertation contributes a Natural Language Generation (NLG) system that generates simple weather reports in Maltese. To our knowledge, this is the first kind of research that consists of an NLG system which produces Maltese texts. This system works in the domain of weather forecasts as it produces texts about two weather parameters, which are wind and temperature, from data which was obtained from the SumTime-METEO corpus (Sripada & Reiter, 2005). However, my focus is about choosing lexical time expressions for time intervals, but the problem of vagueness makes this research challenging. This problem makes it difficult to decide which time expression to choose for a period of time, but a survey was done to help choose the right expressions. Moreover, some generated messages were also evaluated to assess the accuracy of the system's lexical choices. This was done with the use of another survey, which showed that the problem of vagueness was partly reduced. Thus, the NLG system is able to produce linguistic information including the time expressions that can be understandable by many users.