Traitement automatique de l'ambiguïté lexicale en grec ancien
(2012)
Files
these_kindt_4mai2012.pdf
Closed Access - Adobe PDF
- 8.08 MB
Details
- Authors
- Supervisors
- Abstract
- (en) The work presented in this thesis deals with the lexical analysis of texts written in Ancient Greek. It comes after previous developments of an electronic dictionary of Ancient Greek, the Dictionnaire Automatique Grec (DAG), whose lexical data are directly taken from various corpus and not from traditional dictionaries. The DAG is a « wide coverage » dictionary. This means that for each form, all possible analyses are provided. If such data provide a comprehensive analysis of the forms of a text, it also leaves numerous lexical ambiguities, a unique form possibly belonging to several lemmas. The aim is to develop tools allowing the automatic resolution of lexical ambiguities, in order to obtain annotated and fully disambiguated corpora, minimizing the number and size of manual interventions performed by an operator. To achieve this goal, UNITEX, specific software used in the field of Natural Language Processing (NLP), has been adapted to process Ancient Greek texts. Implemented with linguistic data directly inherited from the Research Project in Greek Lexicology (Project), these tools achieve lexical disambiguation of texts by applying contextual rules written by a human expert. All the developments described are original and inedited, both in terms of computerized technology and in gathering numerous linguistic data. They allow now to introduce Ancient Greek amongst the languages already distributed with UNITEX, respecting the standards in use both by the NLP specialists and the Hellenists. Experiments and evaluations are performed on different corpus of patristic texts and historiographical sources of the Byzantine era written in literary koine. Rules can then be applied to texts of different kinds, mostly works written in classical prose during the Hellenistic period. Finally, the Hellenists have now at their disposal new analysis tools, integrated in a single environment (UNITEX), a real « worksplace » able to process lexicaly annotated corpus that can then be interrogated and exploited according to the users’ needs.
- Affiliations
Citations
Kindt, B. (2012). Traitement automatique de l’ambiguïté lexicale en grec ancien. https://hdl.handle.net/2078.5/92928