Towards a text type taxonomy: a functional framework for text analysis and generation

The aim of this paper is to offer a functionally-motivated text typology framework which can serve as a basis for the description, comparison and classification of a wide variety of texts for generation purposes. For this task, we have adopted a socio-linguistic approach which links the study of the linguistic features encountered in texts with our knowledge of the components of the social situation in which the texts with takes place. These components (the variables of field, tenor, and mode) are proposed as the three dimensions of variation which reflect different selections from the experiential, the interpersonal and the textual components of the language code (organized at the semological, the lexicogrammatical and the phonogical/graphical levels). The correlation thus established between the categories of the situation and those of the linguistic system is illustrated parallel network, which together constitute the potential of variation along which texts can be compared and characterized. The predictive properties of the matrix show the usefulness of the method (based on the notion of register as a sort of "interface" between the system of social activity types and the system of language) for automated text generation