Current Position

Currently I’m a post-doctoral researcher affiliated to the CNGL in DCU, Dublin, Ireland. I’m working on the EuroMatixPlus project together with Josef van Genabith.

As part of the EuroMatrixPlus, organised the Fourth MT Marathon, which took place in DCU 25–30 January 2010. The event was a great success and you can check it out at http://MTMarathon2010.info

I wrote my Ph.D. thesis under the supervision of Andy Way and Khalil Sima’an. I graduated on 10.11.2009. The title of my thesis is “Automatic Generation of Parallel Treebanks: An Efficient Unsupervised System”

Current Activities

I am now organising the Second Joint EM+/CNGL Workshop (JEC 2010), titled “Bringing MT to the User: Research on Integrating MT in the Translation Industry”. The workshop will take place on 4 Nov 2010 at AMTA 2010. Click here to visit its web page.

Building a system that integrates TMs and MT using sub-tree alignment.

Future Plans

1. Perform a user study evaluating the usefulness to post-editors of TM-translation mark-up of fragments that need editing.
   

2. Research ways of evaluating the usefulness of SMT training data to TM systems.
   

All the intended work that follows is put on hiatus until further notice:

1. Modify the sub-tree aligner to allow parse forests as input. Work on the output format for the string-to-string case.
   

2. Test the aligner on SMULTRON data.
   

Past Activities

1. May 2009 – June 2009: Finished writing my Ph.D. thesis.
   

2. Feb 2009 – Mar 2009: Cleared some significant bugs in the DOT system. Did translation experiments using training data generated with various configurations of the sub-tree aligner. Used both the HomeCentre and 10K sentence pairs en-de EuroParl data.
   

3. Jan 2009: Attended the MT Marathon 2009, where I released the sub-tree aligner as an open-source tool.
   

4. Sep 2008 – Nov 2008: Finished developing all major features of the sub-tree alignment system.
   

5. Sep 2008: Finished the major redesign of the DOT system and put it on hold for an indefinite period.
   

6. May 2008: I had a paper accepted for CoLing 2008 in Manchester
   

7. Apr 23, 2008: My son was born!!!
   

8. Apr 2008: Successfully transfered to full Ph.D. track.
   

9. Dec 2007 – Apr 2008: Work on parallelising the sub-tree aligner. Added and tested a full-search-based selection algorithm. As a result submitted a paper to CoLing 2008. Development of a fast DOT Goodman-style grammar extraction tool.
   

10. Oct 2007 – Nov 2007: Major reorganisation and optimisation of the sub-tree aligner code – managed to cut ~75% off the execution time. The code needed for string-to-string alignment is already stable and can be used.
    

11. Sep 2007: Fine-tuning of the sub-tree aligner. Attended TMI-07 and MT Summit XI. My colleagues presented papers which we wrote together and they were received very well. Reading statistical papers in order to get up to speed in this area.
    

12. Aug 2007: Got married and had a 2-week honeymoon.
    

13. May 2007 – Jul 2007: Improvements and optimisations to the sub-tree aligner. Preparation of the final versions of the papers. Extensive experiments and testing using the sub-tree aligner. Development of a fast DOP Goodman-style grammar extraction tool.
    

14. Jan 2007 – Apr 2007: Development of a statistical sub-tree aligner. The main goal was to build a language independent system that could be used for the reliable sub-tree alignment of any parallel treebank.
    This work resulted in the submission of two papers in which I am a co-author – to TMI-07 and MT Summit XI. Both papers were accepted. I also attended the first MT Marathon in Edinburgh.
    

15. Aug 2008: Attended CoLing 2008, where I presented my paper “Automatic Generation of Parallel Treebanks.” I got many positive comments on the paper and even some collaboration proposals.
    

16. Jun 2008 – Aug 2008: Major software redesign of our in-house DOT system. Managed to significantly reduce both the memory requirements and the execution time.
    

17. Oct 2006 – Dec 2006: Getting up to speed in the MT field with emphasis on DOT.