Yesterday was the demo to the technical panel of the software I was creating as a research assistant in DLSU. I wasn't part of the team who went to the meeting though, so I wasn't the one who demo'ed it. In a way, I feel I was being irresponsible. Although I did have a short course inadvertently (honest!) scheduled. Could I have moved it? It's possible, but I didn't. Frankly, I was scared.
The project called for the UI integration of 8 NLP(natural language processing)-related undergrad and graduate thesis software. It wasn't really difficult to create an integrated UI in Java. Deciphering the interfaces of each thesis wasn't that much of a problem too. Although the debugging, testing, and training was hell - and that's why I was scared -- because the engines weren't sufficiently trained, and successful translation can only be done on a targeted set of sentence types.
Not that I didn't try. Studying the internal behavior of each software was really difficult. Moreso in determining what training data to feed each one so that each would behave as expected. It was more or less trial and error until I understood the internals of the various code... well, sort of understood. So we settled for targeted training.
Anyway, the technical panel seemed happy and satisfied. I'm glad they understood the difficulties involved.
As a whole, the project was fun, and a breath of fresh air. I learned Java. I contributed something. I learned new things - such as how a lexical functional grammar (LFG) works; and how to add functionality to OpenOffice.org using Java. I got paid. What more could I ask?
There's still one more demo left, this one more difficult since non-technical people will be involved... and they usually expect something "flashy." So my short-term task would be to further improve the possible sentence set. Also, there's still one other subproject left undone, and it's going to be web-based, and hence something new for me again. These tasks are going to extend past my contract, but I don't really care. I'm just excited! :)
Now for the sad part... It would seem there might not be any follow-up projects along this research line. I was hoping to continue studying, tweaking, and even try integrating the design of the various systems - even without pay, just as a pet project. (I really enjoyed it!) I guess I still could, but if it's not going to be useful, I feel it would just be wasted.
Overall, I feel satisfied about the accomplishments over the past year ever since I resigned from Canon. Satisfied in a way I never felt for the past several years. I felt something was jump-started in my life; with this research assistant work capping a long series of learnings. I'm hoping to still volunteer for research work in the future, even after I, God willing, get a new corporate job (my bills are piling!). I hope I was able to deliver a good work for them to reconsider me. ;)
Many thanks to Nats, Borgz, Sol, Badong, and Doc Rachel, for the support during the course of the project. :) Special thanks to Kat Go for the help with two of the engines.
Cheers, to an interesting year, and a more interesting future!
Kampai!
1 comment:
Cool! You got to do something NLP-related, even if it's mostly UI integration and training. I actually wanted our undergraduate thesis to be something related to NLP or Web Development, but our leader said it was too..."easy." He thinks it's something that wouldn't qualify for MOTA. I might do something NLP-related in my masteral thesis, though. I'm still thinking about it. :)
Post a Comment