So, I had applied for research internships for this summer at IBM and Google. I interviewed with IBM this past Tuesday. My interviewer was Kishore. Pretty smart guy all around and very nice to talk to. I was not really that nervous even though I should have been (The man came up with the evaluation metric for the the area I work in - Machine Translation or MT) !) We chatted for about 30 mins about what all I had done in MT and what his interests were and so forth.
He called me back on Thursday and gave me the good news. Well, good and bad, actually. Good news, of course, that I had been selected for the summer internship program. Bad - I won't be working with Kishore's group. I will be working with another group that works on statistical techniques for named-entity tracking and co-reference resolution. Kishore said that there is virtually no distinction between his group and this one, since they use the same tools and techniques, mostly Maximum Entropy techniques. Christof also brought up some good points about this: 1) No one in our research group actually works on this area, so i will be able to provide this unique focus. 2) Since the tools and techniques are the same for both MT and what I would be working on, nothing is actually stopping me from applying the same to the MT problem.
So, all in all I am pretty excited about IBM. It's the T. J. Watson Research Center in Yorktown, NY. Stacy told me that that's only about an hour from the city. So guess where i will be on the weekends ? :-)
Now, about Google, I have an interview next week. Assuming I get in (which is a big assumption, their interviews are supposed to be pretty rigorous), I would be in quite a quandary. All right, let's weigh the pros and cons:
+ I will get to work with Franz Och, one of the stalwarts of MT and the creator of the state-of-the-art MT system.
+ Google = Cool.
+ Get to live in California for 3 whole months, and that too close to San Francisco.
- Probably have to sign a Non-disclosure agreement and will not be able to publish the research that I do or use it for my dissertation.
- New entrant in the area of Natural Language Processing (including MT).
+ Well respected and established research groups in NLP and MT (The first word-alignment based translation models were called IBM-models, for God's sake! )
+ IBM may not be cool but it is a research giant.
+ IBM has always had strong academic ties and they publish a whole lot more than Google. Good chance of publication and of doing real and meaningful research.
+ Close to NYC and also to MD.
- My work would not directly be in MT.
- No Franz Och.
Well, I listed the pros and cons but I still don't have an answer. I guess I will worry about it if and when the time comes.