| |
Garabik R.
A Simple Russian-Slovak Machine Translation System
Presented article describes a machine translation system, based on
expanded morphology mappings between bilingual dictionary entries, using
simple rules to select appropriate grammar category pairs.
Target-language semantic and morphological disambiguation is achieved by
using a 1st or 2nd order Markov chain language model trained on surface
word forms of a big target language corpus. The approach is, therefore,
applicable to a pair of related languages with similar syntactic and
morphological features, and has been used to produce a Russian-Slovak
machine translation system. In addition, genetic similarities in Russian
and Slovak vocabularies are used to improve the translation by applying
a suitable transliteration scheme for Russian words not present in the
dictionary. Dependence of translation quality on several parameters of
training data properties - dictionary size, amount of noise in the
dictionary - has been investigated.
Back |
