So basically I'm training a sequence to sequence model that translates English sentences to Arabic sentences. I'm using the data provided by Anki @ manythings. I realized that some of the sentences in English (source) have multiple sentences in Arabic (target), for example:

This is one case, where the Arabic harakat are not shown but the idea is that the same word has different translations (yes in arabic the first, fourth and fifth are not the same translations).


A better example is the following one:


I'm not sure how to deal with these cases, should I reduce the data and keep one translation, or should I have for each source key a list of target values. Any advice or "tips & tricks" in preparing the data before training translation models?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.