Alignement lexical en corpus comparables : le cas des composés savants et des adjectifs relationnels

TL;DRAbstract

Our work concerns the automatic extraction of a list of aligned terms with their translations (i.e. specialized bilingual lexicon) from comparable corpora belonging to a specific domain. Comparable corpora include texts written in two languages which are not mutual translations but belong to the same domain. This thesis contributes to the improvement of the quality of an extracted bilingual lexicon. We propose methods dedicated to the translation of two types of terms that have common characteristics among many languages or that cause specific problems for translation due to their nature. These types of terms are the neoclassical compounds (terms containing at least one root borrowed from Greek or Latin) and the terms composed of one noun and one relational adjective. We also propose a method that exploits contexts rich in domain-specific terms to re-rank some provided translations in a bilingual lexicon for a given term. The experiments are performed using two specialized comparable c

Chat with Paper

AI Agents for this Paper