Back to list

Detail of contribution

Auteur: Natalia GRABAR

Co-Auteur(s): Marie-Aude LEFER, Institut libre Marie Haps / Université catholique de Louvain, Belgium

Titre:
French evaluative prefixes in translation: from automatic alignment to semantic categorization


Abstract/Résumé: This presentation deals with French evaluative prefixes and aims to find out whether translation can shed light on the semantics of prefixation by adopting Noel’s (2003) ‘translations as evidence for semantics’ approach. In French, evaluative prefixes can be classified along two dimensions (Wierzbicka, 1991; Fradin & Montermini, 2009): quantity dimension with a maximum/minimum axis and the semantic values BIG (increase; abundance) and SMALL (decrease; attenuation; approximation); quality dimension with a positive/negative axis and the values GOOD (excessive degree; higher rank) and BAD (lack; lower rank). Our objective is to provide corpus-based insights into this semantic sub-categorization of morphological evaluation in French. To do so, we analyze evaluative prefixes alongside their English translation equivalents in a parallel corpus, focusing on periphrastic translations, as they are likely to ‘spell out’ the meaning of the French prefixes. The data used were extracted from the Europarl parallel corpus (Koehn, 2005; Cartoni & Meyer, 2012). We first aligned the French prefixed words with the corresponding word(s) in English target sentences with the help of GIZA++ (Och & Ney, 2000) and a tailor-made alignment program, before proceeding to the evaluation of the aligned sequences and the manual analysis of the bilingual data. Results confirm that periphrastic translations can be used as evidence for semantics in morphological research. For example, a range of typical periphrases of the EXCESS meaning have been identified, such as excess(ive) (e.g. hyperflexibilité - excessive flexibility), overly ‘too’ (e.g. sururbanisé - overly built-up), etc., which help characterize the semantics of evaluative prefixes.