Enabling Multi-Source Neural Machine Translation By Concatenating Source Sentences In Multiple Languages
By Raj Dabre, Fabien Cromierès, Sadao Kurohashi. 2017
33 citations recorded by the source corpus.
In this paper, we explore a simple solution to "Multi-Source Neural Machine Translation" (MSNMT) which only relies on preprocessing a N-way multilingual corpus without modifying the Neural Machine Translation (NMT) architecture or training procedure. We simply concatenate the source sentences to form a single long multi-source input sentence while keeping the target side sentence as it is and train an NMT system using this preprocessed corpus. We evaluate our method in resource poor as well as resource rich settings and show its effectiveness (up to 4 BLEU using 2 source languages and up to 6 BLEU using 5 source languages). We also compare against existing methods for MSNMT and show that our solution gives competitive results despite its simplicity. We also provide some insights on how the NMT system leverages multilingual information in such a scenario by visualizing attention.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Marjan Ghazvininejad
- Luke S. Zettlemoyer
- Anoop Kunchukuttan
- Chenhui Chu
- Raj Dabre
- Graham Neubig
- Nabil Hossain
- Arda Tezcan
In Computer Vision and Pattern Recognition
- Anoop Kunchukuttan
- Chenhui Chu
- Marjan Ghazvininejad
- Raj Dabre
- Luke S. Zettlemoyer
- Bram Bulté
- Arda Tezcan
- Nabil Hossain
In Natural Language Processing
- Marjan Ghazvininejad
- Luke S. Zettlemoyer
- Anoop Kunchukuttan
- Chenhui Chu
- Raj Dabre
- Graham Neubig
- Nabil Hossain
- Arda Tezcan
Citations in the corpus
- Simple and Effective Retrieve-Edit-Rerank Text Generation (2020, background): “, 2018; Cao and Xiong, 2018) and post-editing for NMT (Hokamp, 2017; Dabre et al., 2017).”
- A Survey of Multilingual Neural Machine Translation (2019, methodology): “Zoph et al. (2016) and Dabre et al. (2017b) have empirically shown that language relatedness between the parent and child source languages has a big impact on the possible gains from transfer learning.”
- Neural Fuzzy Repair: Integrating Fuzzy Matches into Neural Machine Translation (2019, methodology): “A similar simple concatenation approach has also been used to take advantage of multiple source languages to increase the quality of NMT output (Dabre et al., 2017).”