Findings of the Second Workshop on Neural Machine Translation and Generation

Abstract

This document describes the findings of the Second Workshop on Neural Machine Translation and Generation, held in concert with the annual conference of the Association for Computational Linguistics (ACL 2018). First, we summarize the research trends of papers presented in the proceedings, and note that there is particular interest in linguistic structure, domain adaptation, data augmentation, handling inadequate resources, and analysis of models. Second, we describe the results of the workshop's shared task on efficient neural machine translation (NMT), where participants were tasked with creating NMT systems that are both accurate and efficient.

MOROCCO: Model Resource Comparison Framework

“…The trade-off between model performance and computational efficiency has been explored in multiple shared tasks and competitions. The series of Efficient Neural Machine Translation challenges [1,4,6] measured machine translation inference performance on CPUs and GPUs with standardized training data and hardware. The performance was evaluated by the BLUE score, while the computational efficiency was measured by multiple parameters, including real-time, which the model used to translate the private test set, peak RAM and GPU RAM consumption, size of the model on disk, and the total size of the docker image, which could have included rule-based and hard-coded approaches.…”

OpenNMT System Description for WNMT 2018: 800 words/sec on a single-core CPU

“…As neural machine translation becomes more widely deployed in production environments, it becomes also increasingly important to serve translations models in a way as fast and as memory-efficient as possible, both on dedicated GPU and on standard CPU hardwares. The WN-MT 2018 shared task 1 focused on comparing different systems on both accuracy and computational efficiency (Birch et al, 2018).…”

Proceedings of the 2nd Workshop on Neural Machine Translation and Generation

“…In this study, we introduce the NICT neural translation system at the Second Workshop on Neural Machine Translation and Generation (NMT-2018) (Birch et al, 2018). A characteristic of the system is that translation qualities are improved by introducing self-training, using open-source neural translation systems and defined training data.…”

“…As neural machine translation becomes more widely deployed in production environments, it becomes also increasingly important to serve translations models in a way as fast and as memory-efficient as possible, both on dedicated GPU and on standard CPU hardwares. The WN-MT 2018 shared task 1 focused on comparing different systems on both accuracy and computational efficiency (Birch et al, 2018). This paper describes the entry for the OpenN-MT system to this competition.…”

“…This paper describes the submissions of the "Marian" team to the Workshop on Neural Machine Translation and Generation (WNMT 2018) shared task (Birch et al, 2018). The goal of the task is to build NMT systems on GPUs and CPUs placed on the Pareto Frontier of efficiency in accuracy.…”