The OLST is delighted to announce that one of its members, François Lareau, obtained a three-year research grant from the Fonds de recherche Société et culture (FRQSC) for a project entitled: Processing collocations in natural language generation.
Abstract: Natural language generation consists in producing–dynamically and automatically–texts that express information to be communicated (usually from raw data). Most natural language generation systems have been designed for a specific language and a given application. However, the aim now is to design more generic systems that could be reused in a range of applications and languages. In order to achieve this, the main linguistic phenomena found in most or all languages must be modelled. One fundamental phenomenon, collocations, has received little attention in natural language generation so far. These include idiomatic expressions such as strong taste, intense flavour or burning desire, in which a specific sense (here, intensity) is not expressed with the same linguistic means depending on the noun to which it applies). In lexicography, collocations are modeleld with functions that correspond to recurrent collocational patterns. For instance, the three expressions mentioned above are encoded using a function called Magn: Magn(taste) = strong, Magn(flavour) = intense, Magn(desire) = burning. Hundreds of such functions have been defined. Our proposal consists in integrating them systematically and entirely to a natural language generation platform for English, French and Spanish in order to solve the problem of collocations in generation.