1 Ontological Issues for Modelling Aspect

The CASK method (Computer-aided Acquisition of Semantic Knowledge) is being used at CELTA with the SEMANA (SEMantic ANAlyser) software, which was especially designed for this purpose. We use the notation of semantic feature structures as a meta-language for describing the category of aspect in Polish regardless of its pertaining to various levels of expression (morphological, syntactical or lexical). To model the category of aspect we treat types of situations as conditions for two relevant aspect parameters we call analysis of the situation (or internal aspect) and control of the situation (or external aspect). In order to cope both with the lexical diversity of aspectual morphemes (prefixes) and the grammatical (dichotomous and obligatory) character of aspect in Slavic languages, we have proposed to define the perfective aspect as a hypercategory. Our database of Polish aspect was analysed using KDD statistical tools (G. Sauvet 2008, http://celta.parissorbonne.fr/anasem/papers/). We obtained the first preliminary experimentally fixed semantic definitions of the perfective and imperfective values of the category of aspect in a Slavic language. 1. Ontology and Semantics Multi-lingual contrastive studies need to refer to ontologies as tertium comparationis, including those which were especially designed in order to account for linguistic objects (expressions). We claim that this task cannot but be interactive (computer-aided), if we want to avoid the lack of precision and the variability of semantic parameters in traditional linguistic research, especially hazardous in the semantic domain. We define the semantic content of a linguistic expression as a function mapping this expression on ontological concepts. Therefore, to describe the semantic content of aspectual expressions used in utterances we need to specify the ontology to which they refer. A formal cognitive description aims at giving an ontological account of a semantic category by treating its definition in different languages as a finite set of strictly defined feature structures. We aim at giving a two-fold account of semantic categories by: building a general set of ontological abstract structures necessary to interpret aspect in different languages choosing a specific set of semantic feature structures for the described category in each human language, in our case for Polish. To describe aspectual values of verbs in context, we had first to define relevant aspectual semantic attributes and their values (AV). At this stage, linguists must use both 1 Acknowledgment: the research on aspect with the SEMANA software has been conducted at CELTA (Paris-Sorbonne University) since 2005 by Hélène and André Włodarczyk (ontology and semantics of verbal aspect), Georges Sauvet and André Włodarczyk (conception of the SEMANA software), Georges Sauvet (analysis of the aspect database with SEMANA KDD tools). Doctoral and master students also take part in this research.