I posted this on another list, but I thought it was worth posting here too. Especially since the tool is FLEx-centric.
In some situations in the world, two language varieties are closely related and there is a translation in one variety, but not in the other. The varieties may be similar enough that a few linguistic tweaks and some lexicon mapping is all that is needed to translate one variety to the other.
CARLA was a tool that did that for us. Unfortunately CARLA doesn’t integrate with today’s SIL tools. Also there have been complaints that setting up CARLA is too complex. FLExTrans seeks to solve those problems and improve on CARLA.
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.
FLExTrans is basically a transfer-based machine translation system like CARLA that uses FLEx as a core component. The 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.
4. Run the six FLExTrans modules inside of FLExTools and you’re done.
FLExTrans in many ways is not new piece of software, rather it glues together many great software packages that already exist. FLEx is at its 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. The transfer rules are written in Apertium’s XML format which I’ve templatized using the XMLMind XML Editor in a similar way to how XLingPaper does it. This way you edit rules using a GUI instead of text editor.
You may ask has this system been used in the field? It is still new and experimental, but Roger Stone in the Philippines did a proof-of-concept test of FLExTrans using two Philippine languages and presented it at a conference recently. See his paper, Implementing Lexical Based Machine Translation Using FLExTrans. His test was successful and he has positive things to say about FLExTrans.
I’m available to help projects that want to experiment with FLExTrans.
One limitation of the tool right now is that it is currently a Windows-only tool. It also does not have the option yet to use HermitCrab to synthesize the target text (I would be interested to know if this is critical for you).
Get more information by going to the FLExTrans site on Github. It is open-source software.
Ron Lockwood
--
You are subscribed to the publicly accessible group "FLEx list".
Only members can post but anyone can view messages on the website.
---
You received this message because you are subscribed to the Google Groups "FLEx list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to flex-list+...@googlegroups.com.
To post to this group, send email to flex...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/flex-list/018201d2352c%243e16f210%24ba44d630%24%40sil.org.
For more options, visit https://groups.google.com/d/optout.