Dean Völgyi (2010) Machine translation based on comparison of text features. EngD thesis.
Abstract
The goal of the thesis was to study the feasability of translating between two natural languages exclusively on the characteristics of those languages, i.e. without using a dictionary. The term 'characteristics' as used in the study signifies the absolute and the relative frequencies of words, the context where these words are found in, as well as the frequency of bi-grams and tri-grams within the subject of a given text. For that purpose we have developed a translation system, into which we have implemened different approaches to translation. The success of the application was evaluated by a Slovenian-English pair of texts. We established that such way of translating is possible to a certain level, however, the results still leave much room for improvement so that this approach becomes useful for more serious purposes, as well.
Actions (login required)