Welcome to Ncode, an open source statistical machine translation system based on bilingual n-grams.

In this site you will find the instructions to download and install the toolkit as well as a sample demonstrating how to use it. Additionally, you will find information considering different of its aspects: its particular translation units, descripton of the implemented feature functions and details about its training, decoding and optimization process.

Note that in order to run Ncode, some additional toolkits are required (i.e. need to be installed in your system):

• kenlm
• OpenFST
• SRILM
• Moses (MERT)

You can see the slides used to present Ncode at the MT Marathon 2011 in Trento (Italy).

Acknowledgements

The system has been developed by Josep M. Crego, helped with the initial ideas of design, implementation suggestions and testing the software by: Adrià de Gispert, Marta Ruiz, Patrik Lambert and José B. Mariño (UPC, Barcelona) and Alexandre Allauzen, Aurélien Max, Thomas Lavergne, Artem Sokolov and François Yvon (LIMSI, Orsay).




Mainly funded by the TC-Star (2005-2008) and Quaero (2008-2011) projects.