Linguistic expressions as a tool to extract information

When searching for multiple words in a corpus database, we all have encountered the problem of not coming up with ‘the right thing’. Indeed, very often the results produced by the computer do not meet the analysis we would have produced as humans. In a text, in order to understand it correctly, we rely not only on the semantic content of the words but also on more grammatical means, which provide reading instructions. Those are the ones we will be dealing with here and we will use French prepositions (pour ce qui est de, en ce qui concerne, à propos de, au sujet de, etc.) which have the syntactic properties of being either detached and generally in an initial position or, of being dependent on another morphosyntactic constituent. This syntactic behaviour is of importance in the way in which information is processed and corresponds to two discoursal functions. When detached, these prepositions introduce a thematic (Charolles 1997) and mark an author’s pragmatic intentions. When introducing a thematic they can integrate one or more propositions. When they depend on another constituent they simply focalise and do not have that integrative property. The ultimate aim of this ongoing analysis is to engineer a tool using linguistic expressions to extract thematic information. The purpose of this paper is three fold: to detail how linguistic information concerning the thematic introducers is captured; to show how they help in segmenting texts; and to give a technical overview of the system currently being developed by the LaLIC team, utilising this linguistic information.