Hi everyone,
today we have had a meeting around this year's GSoC and we would like to
propose a task which is described below. Any suggestions are welcome,
specially those, if any, that find potential overlaps with other Apertium's
ongoing development lines.
*Possible titles: (we prefer the first over the second one)*
*a) Extracting knowledge from Apertium's post-edition logs to improve
translations *
*b) Knowledge mining over Apertium's post-edition logs*
*c) (anything more appealing and meaningful you could think of)*
*Difficulty:*
Entry level / Medium
*
*
*How? (required skills)*
XML, PHP, Java
*What? (description)*
The goal of this task is to build the means to enable the mining of the pre
and post edition interface log files in order to extract information that
could be incorporated into Apertium's engine in the form of rules or
dictionary entries.
*Why? (rationale)*
Apertium has a pre and post edition interface that allows users to correct
errors in original documents (before translating it) and also in the
translation offered by Apertium. The interface integrates different tools
like spell and grammar checkers and integrated dictionary look up. The
changes that a user performs in this pre and post edition interface are
very valuable since they are human corrections on top of Apertium's output.
These changes are logged in and represent a very rich information source in
order to further improve the engine's performance.
*
*
*Who? (mentors)*
Luis Villarejo
Jordi Duran
Jimmy O'Regan
and maybe Arnaud Vié or Camille (students involved in the pre and post
edition interface development)
Let us know your feedback on it.
Best,
Luis
------------------------------------------------------------------------------
Keep Your Developer Skills Current with LearnDevNow!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-d2d
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff
