FLExTrans is a transfer-based machine translation system similar to CARLA that uses Fieldworks Language Explorer (FLEx) as a core component. A source text starts out in the source language’s FLEx project and ends up after the translation process in the target language’s FLEx Project. The basic steps are as follows:
1.Analyze the text in the source FLEx project.
2.Map lexical senses that are in the source text to senses in the target FLEx project (there’s a Sense Linker Tool to help with this).
3.Write transfer rules that convert source words and phrases to target words and phrases.
FLExTrans in many ways is not a new piece of software, rather it glues together many great software packages that already exist. FLEx is at the core, but it also uses STAMP for synthesis and an open-source engine called Apertium for the transfer component. Apertium’s engine is very powerful and it's in Apertium’s transfer XML format that rules are written. I’ve templatized the format using the XMLMind XML Editor in a similar way to how XLingPaper does it. In this way you edit rules using a GUI instead of text editor.
FLExTrans came out of my research while doing a Master of Science in Computational Linguistics through the University of Washington in 2015. See my thesis entitled A Linguist-Friendly Machine Translation System for Low-Resource Languages.
|Support||free, email author, forum or wiki|
|License||GPL (GNU Public License)|
|Download page||FLExTrans Download|