Moses
statistical
machine translation
system

Ergun Bicici's Parallel FDA5

Ergun has made available his Parallel FDA5 dataset for download.

   https://github.com/bicici/ParFDA5WMT
   https://drive.google.com/folderview?id=0B6Jae6trZb1aVzJvSnZ5UncxTEE&usp=sharing

Models trained on Europarl

The MosesCore team makes available all the models they train from Europarl data. This includes phrase-based models, hierarchical models, and factored models.

Also available for download are the Moses source code and the (64-bit Linux) executables that was actually used to run the experiments.

This is provided this as part testing and release of the Moses toolkit.

If you have any queries or question, please contact Hieu Hoang or Barry Haddow.

Phrase Tables

Used to translate short sections of the source sentence into target sentence.

Moses can read text files formatted like Pharaoh phrase tables, or it can read formatted binary files. However, binary file as loaded on demand for each sentence so decoding is quicker and requires less memory. Therefore, we have provided binary files below.

Language Model

Give a score to the 'correctness' of the target sentence

Small Samples

Edit - History - Print
Page last modified on September 06, 2014, at 08:58 AM