Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Design of an Open-Source Architecture for Neural Machine Translation

Séamus Lankford, Haithem Afli, Andy Way

TL;DR

A single-click model development approach has been implemented, and models developed by adaptNMT can be evaluated using a range of metrics, and the application provides an intuitive user interface that facilitates hyperparameter customization.

Abstract

adaptNMT is an open-source application that offers a streamlined approach to the development and deployment of Recurrent Neural Networks and Transformer models. This application is built upon the widely-adopted OpenNMT ecosystem, and is particularly useful for new entrants to the field, as it simplifies the setup of the development environment and creation of train, validation, and test splits. The application offers a graphing feature that illustrates the progress of model training, and employs SentencePiece for creating subword segmentation models. Furthermore, the application provides an intuitive user interface that facilitates hyperparameter customization. Notably, a single-click model development approach has been implemented, and models developed by adaptNMT can be evaluated using a range of metrics. To encourage eco-friendly research, adaptNMT incorporates a green report that flags the power consumption and kgCO${_2}$ emissions generated during model development. The application is freely available.

Design of an Open-Source Architecture for Neural Machine Translation

TL;DR

A single-click model development approach has been implemented, and models developed by adaptNMT can be evaluated using a range of metrics, and the application provides an intuitive user interface that facilitates hyperparameter customization.

Abstract

adaptNMT is an open-source application that offers a streamlined approach to the development and deployment of Recurrent Neural Networks and Transformer models. This application is built upon the widely-adopted OpenNMT ecosystem, and is particularly useful for new entrants to the field, as it simplifies the setup of the development environment and creation of train, validation, and test splits. The application offers a graphing feature that illustrates the progress of model training, and employs SentencePiece for creating subword segmentation models. Furthermore, the application provides an intuitive user interface that facilitates hyperparameter customization. Notably, a single-click model development approach has been implemented, and models developed by adaptNMT can be evaluated using a range of metrics. To encourage eco-friendly research, adaptNMT incorporates a green report that flags the power consumption and kgCO emissions generated during model development. The application is freely available.
Paper Structure (15 sections, 1 figure)

This paper contains 15 sections, 1 figure.

Table of Contents

  1. Credits
  2. Introduction
  3. Related Work
  4. NMT
  5. NMT Tools
  6. Architecture of adaptNMT
  7. adaptNMT
  8. Initialization and logging
  9. Modes of operation
  10. Customization of models
  11. Use of subword segmentation
  12. Translation and evaluation
  13. Infrastructure
  14. Discussion
  15. Conclusion and Future Work

Figures (1)

  • Figure 1: Proposed architecture for adaptNMT: a language-agnostic NMT development environment. The system is designed to run either in the cloud or using local infrastructure. Models are trained using parallel corpora. Visualization and extensive logging enable real-time monitoring. Models are developed using vanilla RNN-based NMT, Transformer-based approaches or transfer learning using a fine-tuning approach. Translation and evaluation can be carried out using either single models or ensembles.