Week 9: August 12 - 16
This week was a relatively boring week for me as I continued my work on the statistical part of ELAN. I studied the various contingency table formulas and discussed them with the team to find out which ones we would need. After some more research, I gave up on determining the exact formulas and tried to simply implement as many as I could within the limited time frame! That way, I could defer the "hard" work of figuring out which metric to utilize on our corpus to a future date. As usual, it was fun to continue hacking away at the ELAN codebase and work on the code instead of wading through the dense publications. I'm majoring in mathematics at my university but some of the publications are really complex and build on so much previous work that it would take me a while to fully understand the implications of the formulas. Fortunately, like I mentioned in the previous week, the NSP package was a big help. It was easier for me to look at code instead of publications and from there it was almost straightforward to port the Perl code over to ELAN's Java codebase.
Other than the usual statistics work, I also finished up the Tapia conference submission with my co-author and polished it a bit more. It was gratifying to have the extra time to make sure that we didn't miss anything and decide what to cram into a measly 2 page paper, ha! It felt so good to submit yet another publication, wow! I feel like I belong in academia now, with several papers now under my belt. It's hard to believe what one can accomplish during the summer and I have never been this busy during a period of time that is normally vacation time for me. Here's the details of our Tapia paper:
- Title: Exploring Corpus Analysis for Coarticulation in ASL
- Authors: Larwan Berke and Marie Stumbo
- Abstract: The American Sign Language (ASL) Project at DePaul University is developing an automatic English to ASL translator. A survey of linguistic literature focused a decision to improve coarticulation support in order to generate more natural movement and better convey the nuances of ASL. Statistical tools have been developed to provide a more detailed analysis of timing from annotations in an ASL corpus. The resulting data will be used to create more accurate animations.
Once we were done with the Tapia paper, the only other thing I had to do was to finish my ELAN-related work before the summer is over! I talked with my mentor and we decided that I would wrap up the internship in the final week, which is next week. One of the biggest tasks for me to do is to finalize the ELAN code and stop hacking on it! When I finish my work on the code, I can start writing the final report for DREU and if I keep hacking on the code I will end up revising the paper too much. The final report's requirements state that it should be a technical paper suitable for a conference but I thought it was a trivial thing to complete as I've already practiced a lot over the summer! I can't believe it's almost over and that meant it was time for me to start preparing for the fall semester at Gallaudet University.