I write a lot of short emails giving advice to people thinking about graduate study. Most of these are written in response to specific questions, but I've started to save the ones of more general interest here.
This page contains my own personal suggestions about how to persue graduate study. This is free advice, and I make no guarantees about the accuracy, efficacy or completeness of the information and recommendations presented here. Graduate study is much more varied and individual than undergraduate study, and you should check and confirm the information, advice and recommendations contained here yourself, of course.
In general your complete applications should be at the graduate school by the end of the year preceding the year you hope to be admitted (this may vary from institution to institution, so you should check, of course). Since the lead time for GREs and the like may be many months, you need to start preparing your application early in the year before which you plan to start graduate school.
Most top graduate schools strive to admit the best applicants they can find, and Web-based application procedures make it easier to apply to US graduate schools from abroad. But non-US students should recognize that visa requirements for students are now considerably more stringent than they were just a few years ago.
You should pick the schools you apply to carefully. At the graduate level you'll be working closely with one or two professors and each professor's areas of expertise are different. Moreover, research is largely driven by hunches about where the biggest advances in the field will come, and the professors' hunches will determine what they and their students do. So my advice is: do your homework before you pick the schools you're going to apply to; go to professors' web pages, download and read their papers and try to find about 4-7 professors whose research direction you really like, and apply to their departments. It may also be wise to contact the professors directly and let them know you are applying.
I think every applicant should try to imagine what it's like for the admissions committee in an active research department. Every good department will get far more applicants than they have positions available, so they will select the applicants that are most likely to succeed in doing innovative research and completing their PhD.
Putting it very crassly, there's a huge difference between undergraduate and graduate education in the US. Undergrads pay for the privilege of attending college, and your education and edification is the main goal of the undergraduate experience. But as a graduate student in fields like computational linguistics and cognitive science you can generally expect to be fully supported financially while you are studying, i.e., you are paid to study, and the relationship between student and institution changes accordingly. While we hope that you'll be personally enriched and enlightened by your graduate education, this is not the reason why you're being paid to study. Instead, we expect that you'll advance scientific knowledge to the benefit of all society. Of course you may have personal goals for going to graduate school and it's fine to mention those in your application, but your primary goal in your application should be to demonstrate that as a PhD student you will contribute to and advance our scientific understanding of the field you're studying.
In general, faculty want graduate students that can do ground-breaking research and complete a PhD in a timely fashion in the research areas they are interested in. How do you convince the faculty that you can do this? Well, first of all you need to demonstrate that you understand the field well enough to know what you're getting into. It's a waste of everybody's time if a student decides after a year or so that they don't really want to work in the field they have been admitted to study. More than that, you need to demonstrate that you are honestly interested in the field you are applying to study. There are easier ways of making a living than as a research scientist (think of the ``1 per cent inspiration, 99 per cent perspiration'' Edison quote) so we're looking for someone that's sufficiently inspired and motivated to actually stick at the research long enough to get something substantial done.
Second, you have to demonstrate that you have the technical ability to work in the field of your choice. Notice that I said ability; you don't need to demonstrate that you already have all of the skills you'll need, as you'll get training in graduate school. But again, it's a waste of everyone's time to admit a graduate student and discover a year or so later that they can't program a computer or do statistics or whatever, so if skills like those are important to your field, you definitely need to demonstrate that you have the ability to learn those skills. And a very good way to do this is to take a couple of courses in these areas and do well in them. (It doesn't hurt to have a professor from these areas write a letter of recommendation for you). So you should make clear in your application that you have the skills, or the ability to learn the skills, that you'll need to do your research.
Finally, you should have a moderately specific idea of the kind of research you want to do. It should fit in well with the research program of at least one of the faculty members in the department you're applying to (I discuss this elsewhere on this page). Your interests should be neither too vague (you should know the field well enough to know what the important issues in the field are) nor too specific (as you are coming to learn from and work with faculty).
These days computational linguistics draws on computer science, linguistics and statistics, and good computational linguists are reasonably proficient in all of these. You don't need to be an expert in all of these areas (after all, you will be trained in graduate school), but you should try to take enough substantive courses in each of these areas to demonstrate that you have ability in these areas. And if your university or college offers a computational linguistics course, you should definitely take it.
There's no fixed set of prerequisites required to apply to a computational linguistics program. It's hard to imagine anyone doing research in computational linguistics without having some programming skills (this is often the greatest hurdle for linguists that would like to become computational linguists). These days mathematical skills (say, in statistics and probability theory) are also very important. You should also take one or two courses in linguistics (say, syntax) or computational linguistics in order to demonstrate that you have some idea of what the field is about. My own advice is: get as much training as you can in the mathematical and computational sciences while an undergraduate!
This is a hard question. It sometimes happens that students only discover computational linguistics in their last year at college, and then it is too late to take courses in computer science, linguistics and statistics as I suggested above. My general advice is: apply for graduate school this year, but if you don't get accepted, try to spend the year getting experience that will strengthen your chances of being accepted next year. It's ideal if you can get a job as a research assistant in a company that is doing real computational linguistics research, but beware of jobs that may be called computational linguistics but really only involve fairly mundane data entry.
It is possible to take courses at community colleges or other part-time post-graduate institutions; this is a way of getting background in subjects that you should have covered as an undergraduate, but didn't.
Computational linguistics is an interdisciplinary subject, and these days most computational linguistics research is done in computer science departments, but linguistics departments are also increasingly offering computational linguistics as well. To a lesser extent, we're also seeing computational linguistics research done in applied mathematics and statistics departments.
Brown University exemplifies this trend, with computational linguistics research and related work being done in our Computer Science department by Professor Charniak, by me (Mark Johnson) in the Cognitive and Linguistic Science department, and by Professor Geman in the Department of Applied Mathematics.
Each of these faculty have their own specific research interests, of course, but they follow broad disciplinary interests. Overgeneralizing somewhat, computational linguistics research in computer science tends to focus on algorithms and implementations and models that can be efficiently implemented on modern machines. Computational linguistics research in linguistics tends to focus on and draw more on linguistic theory and modeling human performance (say, as in computational psycholinguistics). Computational linguistics research in applied maths focuses more on the mathematical aspects of models and investigating the mathematical properties of grammars.
It probably makes sense to apply to the department that best matches your own research interests and skills. In fact, I think that graduate school is much more intensely personal than undergraduate eduation is, and you should pick the departments you apply to largely on the basis of the specific people that you might work with should you wind up going there. It's hard work, but you should study the recent research produced by the faculty and students at all of the departments you're thinking of applying to, as that will give you a good idea of what you could be doing if you were to go there. In addition, having a good understanding of the research done in a department will enable you to prepare a much better graduate school application, so it is time well spent.
5th November 2006