Moses
statistical
machine translation
system
Moses statistical machine translation system |
1. Moses
2. Getting Started
3. Tutorials
4. Training
- Overview
- Prepare training data
- Factored Training
- 1 Prepare data
- 2 Run GIZA
- 3 Align words
- 4 Lexical translation
- 5 Extract phrases
- 6 Score phrases
- 7 Reordering model
- 8 Generation model
- 9 Configuration file
Language Models
- Tuning
- Training Reference
- Decoder Reference
5. User Documentation
- Advanced Models
- Efficient Phrase and Rule Storage
- Search (Decoding)
- Unknown Words
- Hybrid Translation
- Moses as a Service
- Incremental Training
- Domain Adaptation
- Constrained Decoding
- Cache-based Models
- Sparse features
- Support Tools
- External Tools
- Web Translation
- Moses2
- Pipeline Creation Language
- Obsolete Features
6. Development
- Video
- Code Guide
- Code Style
- Factors
- Feature Functions
- Sparse Feature Functions
- Code Documentation
- Regression testing
7. Background
Welcome to Moses!
Moses is a statistical machine translation system that allows you to automatically train translation models for any language pair. All you need is a collection of translated texts (parallel corpus). Once you have a trained model, an efficient search algorithm quickly finds the highest probability translation among the exponential number of choices.
Moses on Twitter
Tweets by @MosesSMTNews
- 5 October 2017 Moses v 4.0 has been released!
- 8 September 2016 Moses2, a fast drop-in replacement for the Moses decoder
- 12 December 2015 Add a new feature function to Moses
- 17 June 2015 Slate for Windows
- 15 June 2015 Moses, and more, on Amazon cloud Box
- 1 June 2015 Developing Moses with Eclipse video
- 4 February 2015 Moses v 3.0 has been released!
- 21 July 2014 Moses now has nightly speed tests
- 14 July 2014 How to compile Moses with Eclipse
- 4 March 2014 Bug fix release for Moses, now version 2.1.1
- 3 February The 2014 Machine Translation Marathon will take place in Trento, Italy from 8-13th September.
- 21 January 2014 Moses v 2.1 has been released!
- 26 March 2013 The 2013 Machine Translation Marathon (MTM2013) will take place in Prague, Czech Republic from 9-14th September
- 5 March 2013 What do you want to see in Moses v2.0? See here for projects and how to suggest them.
- 28 January 2013 Moses v 1.0 has been released!
- 12 October 2012 Moses v 0.91 released
- February 2012: Moses development is being supported by the EU under the MosesCore project
- September 2011: Moses now has a cruise control page to see the status of the current builds
- September 2011: Moses is now hosted on github
Features
- Moses offers two types of translation models: phrase-based and tree-based
- Moses features factored translation models, which enable the integration linguistic and other information at the word level
- Moses allows the decoding of confusion networks and word lattices, enabling easy integration with ambiguous upstream tools, such as automatic speech recognizers or morphological analyzers
- The Experiment Management System makes using Moses much easier
Get started
The released software includes a command line executable which can used for decoding. The source code for the decoder, can be downloaded from github. Download the latest release or the current snapshot from github.
Learn about the decoder, training models, and tuning. Follow the step-by-step guide to build a baseline translation system. The documentation available at this web side is also compiled in a printable manual.
Acknowledgement
The development of Moses is mainly supported by the European Union under the following projects:
- EuroMatrix and TC-STAR (Framework 6)
- EuroMatrixPlus, LetsMT, META-NET, MosesCore and MateCat (Framework 7)
It has received additional support from
- University of Edinburgh, Scotland
- Charles University, Prague, Czech Republic
- Fondazione Bruno Kessler, Trento, Italy
- RWTH Aachen, Germany
- University of Maryland, College Park, United States
- Massachusetts Institute of Technology, United States
- US funding agencies DARPA, NSF, and Department of Defence
Open Source License
Moses is licensed under the LGPL.