mmT5: Modular Multilingual Pre-Training Solves Source Language Hallucinations.
Jonas PfeifferFrancesco PiccinnoMassimo NicosiaXinyi WangMachel ReidSebastian RuderPublished in: EMNLP (Findings) (2023)
Keyphrases
- cross lingual
- source language
- target language
- cross language information retrieval
- parallel corpus
- machine translation
- cross language
- query translation
- language independent
- machine translation system
- sentence pairs
- feature selection
- statistical machine translation
- document retrieval
- document collections
- active learning
- digital libraries
- bayesian networks
- multimedia