[GSOC Progress] Entity Disambiguation

[chalitha udara Perera <ch...@gmail.com>]

Hi all,

This is my current progress with GSOC work. After discussing
with my mentor Rafa, we have decided to use aidalight[1] as the
disambiguation service to use. Aidalight can be directly integrated
with Stanbol because it is licensed with  Apache licence 2,0.

I have first converted the aidalight project into a maven project and
implemented REST services that can use externally defined mentions
for disambiguation. Therefore it is possible to use stanbol text annotations
with this service.

Code for Aidalight REST server can be found here [2]. Aidalight server
needs a server machine with at least 30GB main memory to run. I am very
grateful for
Rafa for finding me a server that can run aidalight server. We have tested
the server and here is a sample result.

Text : With United, Beckham won the Premier League title 6 times
Manually tagged mentions: United, Beckham, Premier League

Output from Server:
mention: Beckham, entity: David_Beckham, Confidence: 1.120247486075907
mention: United, entity: United_Kingdom, Confidence: 0.6374717182561688
mention: Premier League, entity: Premier_League, Confidence:
1.1851138632800318

I have also implemented an enhancement engine to use aidalight server
with stanbol text annotations. Code for engine can be found here [3].
I am currently testing the engine.

In the remaining time of the project, I will continue improving the code
and completing the documentation. Any suggestions and comments are
more than welcome

cheers,
Chalitha



[1] https://code.google.com/p/aidalight/
[2]
https://github.com/ChalithaUdara/Apache-Stanbol-Disambiguation-API/tree/master/stanbol-aidalight-server
[3]
https://github.com/ChalithaUdara/Apache-Stanbol-Disambiguation-API/tree/master/aidalight-disambiguation



-- 
J.M Chalitha Udara Perera

*Department of Computer Science and Engineering,*
*University of Moratuwa,*
*Sri Lanka*