Latin Translation with Machine Learning: Interpretes Linguae Latinae
Susan Shin, Summer Zhou, Zoe Zitzewitz
Abstract!
This paper presents a neural translation model for Latin. Given the lack of accurate translations and resources in accessible corpora, Latin is arguably one of the most difficult languages for which to implement an accurate automated translation model. However, by finding vocabulary, phrases, and sentences with different complexity levels ourselves and cleaning it both manually and through programming,our model has achieved 96% accuracy in Latin translation. Our experimental results show that the proposed model performs better on vocabulary and short phrases, but not as well for long, complex sentences, and thereby achieves a noticeable BLEU score improvement on Latin-English translation tasks compared to the Google Phrase-Based Neural Machine Translation model.
Github Link
Our Code!Our Poster
.png)