Moses

Moses is a translation tool that allows users to train the computer program to translate between a specific "language pair" (the two languages being translated) by inputting a collection of translated texts into a software program.  The program then analyzes the content of the documents and is able to translate new documents based on statistical reasoning where it inputs the most likely translation of the text based on the usage from the translated source documents. 

What is it used for?

As a statistical translation tool, Moses is used for translating content when there is a large corpus (a collection of sentences and their direct translations into the target language) that can be pulled in the system.  However, even if the corpus of the languages being translated is small, Moses is still able to provide statistical translation as long as there is some data to work with.

Moses features include:

• Beam-search: An effective kind of search algorithm
  
• Phrase-based functionality: Allows translation of text segments and chunks
  
• Factored recognition: Recognizes different "factors" of language (e.g. part-of-speech, morphology)
  

Resources

For more information on Moses, check out these resources:

• Moses Hope Page: http://www.statmt.org/moses/index.php?n=Main.HomePage
• Moses Training: http://www.statmt.org/moses/?n=FactoredTraining.HomePage
• Moses Manual PDF: http://www.statmt.org/moses/manual/manual.pdf 

Moses is free software licensed under the LGPL.