About this document
Hindi-Marathi NMT for WMT 2020 by Krishna Gupta is a document available to read on EtoBox.
This paper describes experiments on neural machine translation for the similar language pair of Hindi and Marathi at WMT 2020. The authors used an attention-based sequence-to-sequence model with morphological segmentation and back translation. They explored using part-of-speech tags and morphological features to help the model learn from the morphologically rich Hindi and Marathi languages. The best results were obtained by combining these techniques with byte-pair encoding for open vocabulary translation.
- Author
- Krishna Gupta
- Language
- EN