Μηχανική Μετάφραση και Μέσα Κοινωνικής Δικτύωσης: Συγκριτική αξιολόγηση μηχανικά μεταφρασμένων κειμένων με χρήση αυτόματων μετρικών
Date Issued
March 30, 2021
Type
Μεταπτυχιακή Διπλωματική Εργασία
Abstract
The main purpose of this master’s dissertation is the study of the relationship between machine translation (MT) and the social media, through the comparative evaluation of machine translated texts, with regard to the language used in the social networking websites and the social media most popular topics. In order to achieve this, a certain number of texts were selected and translated making use exclusively of machine translation software. The machine translated texts were later evaluated applying both automated and manual evaluation methods. In order to be eligible for selection, the texts had either to be extracted from social networking websites or to regard popular topics on the social media, while also being part of bilingual parallel corpora so that they could be evaluated utilizing automated MT evaluation metrics.
Three language pairs were studied: German to English, Italian to English and Italian to Greek. All texts were automatically translated twice, making use of Google Translate and Facebook Translate respectively, while the MTUOC tool (and more specifically MTUOC-Eval) was used to calculate the automated MT evaluation metrics. Specifically to Greek as a target language, a variant of the general tokenizer of MTUOC-Eval is proposed, especially designed for the Greek language.
Subjects