'Lexical Incompleteness in Machine Translation: a multilingual
morphosemantic approach'
The main objective of this research is to evaluate the exploitation of
inference between constructed neologisms in different languages, in the
context of machine translation. In the theoretical aspect of this work we
investigate the grounds for inferring neologisms between languages, and
propose a first attempt at modelling the inference through bilingual Lexeme
formation rules (LFR). To formalise these bilingual LFR, we propose a
methodology grounded in contrastive morphology, which allows us to compare
the morphological systems of the two languages chosen for this experiment
(Italian and French).
In the practical aspect of the work, we describe how we design and
implement a system that automatically translates prefixed neologisms from
Italian into French. Since such an implementation has to be able to deal
with numerous practical problems, especially those stemming from formal
ambiguities and the availability of adequate and appropriate resources,
each step of the translation process is evaluated and refined according to
the available resources. Finally, more general questions of feasibility and
portability are addressed, especially those concerning the possibility of
extending the system to other languages and to other morphological
construction processes.
|
E-mail this message to a friend
