User: Password:
|
|
Subscribe / Log in / New account

Apertium: An open source translation engine

Apertium: An open source translation engine

Posted Dec 16, 2011 7:46 UTC (Fri) by kleptog (subscriber, #1183)
Parent article: Apertium: An open source translation engine

Thank for the article. Sounds like a really cool program. I've always been interested in languages and this combines it with programming.

One thing I'm surprised about is that there's no mention of using existing translations to help constructing new ones. For example, online there are masses of documents translated into various languages. Open source projects are full of PO files. Translated books have are also online. Surely it must be possible to use these to provide hints to missing words in the dictionary?


(Log in to post comments)

Apertium: An open source translation engine

Posted Dec 16, 2011 16:02 UTC (Fri) by n8willis (subscriber, #43041) [Link]

That would be Statistical or example-based MT. The approaches are different enough that they don't mix into a single algorithm -- although a complete MT tool could certainly incorporate more than one method, then somehow attempt to merge the results.

Nate

Apertium: An open source translation engine

Posted Dec 16, 2011 23:38 UTC (Fri) by jimregan (guest, #81866) [Link]

Combining SMT, EBMT and RBMT is quite an active area of research at the moment. It goes by a number of names, but 'Hybrid MT' is the most popular. Hybrid systems which used Apertium as one of their components did quite well in the WMT evaluations this year and last.

Mikel Forcada (the instigator of the Apertium project) maintains a list of open source MT systems of all kinds here: http://www.computing.dcu.ie/~mforcada/fosmt.html

Apertium: An open source translation engine

Posted Dec 16, 2011 23:27 UTC (Fri) by jimregan (guest, #81866) [Link]

Apertium is a rule-based machine translation system, so we don't directly use existing translations. We do, however, indirectly use them to build dictionaries and rules.


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds