The understanding of Shakespeare is a popular topic, especially with English students. However, it can prove difficult to capture its meaning. That's why Svyatoslav and Éléa have tried to implement a two-way modern to Shakespearean English translator. The translations are, alas, not very close, so the end product is a Shakespearean text generator. Their final model got a BLEU score of about 46 and generates somewhat related Shakespearean sentences from a modern English source. It was implemented following a transformer encoder-decoder with attention model trained on sentence pairs scraped from 21 plays of SparkNotes' No Fear Shakespeare side-by-side translations. That is, approximately 60,000 lines of Shakespearean text and the corresponding 60,000 modern equivalents.
Log in or sign up for Devpost to join the conversation.