Do I have a shot at MS in CompLing?

[tlaplace]

I just finished my 3rd year as a statistics major. I've become really interested in computational linguistics, especially probabilistic/statistical approaches to NLP. I'm thinking about applying to MS programs in Computational Linguistics, such as UW or Brandeis, to prepare for getting some sort of NLP job (I don't think my background would be enough to get one upon graduating). However, since I don't have much of an educational background in either CS or Linguistics, I'm worried about whether I have a shot at an MS. 

Overall I have a 3.8 GPA. As far as my math background, I've taken and gotten A's in: Calc 1-3 , Linear Algebra, Probability, Math Stat 1 and 2. Regarding my linguistics/language background, I've taken and gotten A's in: Intro Linguistics, Symbolic Logic, Deductive Systems, Model Theory, Philosophy of Language, and French 1-4.  However, my CS background is lackluster; I've taken: Discrete Math (A-), Programming 1 (B+) and Programming 2 (B+).

I can't fit in any more CS courses before graduating and am not in a position to get rec letters from any CS professors. However, I could get strong letters from statistics professors and the philosophy professors who taught me logic. Also, despite my programming grades I feel competent with Java, R, and Python and could compile a profile of (relatively basic) projects I've completed in those languages.

So, I do feel prepared for an MS in CompLing but am not sure that my background clearly shows it. Do I have a shot? What could I do to strengthen my application?

Edited May 7, 2018 by tlaplace

[ides-of-july]

From my reading of their program materials, the MS in CL program at UW really wants you to have data structures behind you before starting.  It's not a hard and fast requirement (you can take it on deficiency once you're in, if I understand correctly), but it seems understood that your programming acumen should be at or around that level coming in.  So... what did your programming II class cover?  (setting aside questions about grades, for now)

If I read the Brandeis materials properly, it seems like they might be a bit more floppy--but it also seems the calibre of admittees has gone up a fair bit over the last few years, so even though you might clear the basic admissions bar *perhaps* a perception of low skills would hurt.

Note that there are various posts peppered throughout this forum from insiders* who can speak to all of this better, plus you can review the latest materials online from the programs--both are pretty clear about these matters in general.  And I certainly defer to those people here with more knowledge than I have, who should chime in.

Cheers, IOJ
________________________
* Retooling for CL is my general direction as well, but my background is different (anthropology + wandered into the tech industry some years ago)--so I'm going to try doing a standalone linguistics masters, address my deficiencies en passant, and hope that works for the next stage.

Edited May 11, 2018 by ides-of-july

[Gnome Chomsky 2.0]

I graduated from UW CLMS program a few years ago. I had a BA in linguistics and a few CS classes including Data Structures and a probability/statistics class. I completed the MS in one year and did the internship option instead of the thesis.

 

You can complete the program in one year once you start the core classes. A lot of people spent a year taking prereqs in CS and probability/statistics. They're pretty lenient about probability/statistics and will let you self-learn it if you can pass the placement exam. But you pretty much have to take up until Data Structures. You can do this at a community college or at UW. UW has a very well respected CS program and teach Java. You can still take the linguistics and elective CLMS classes in the first year while taking the CS prereqs.

 

As far as the core classes go, it's a lot of programming. There's a programming assignment each week. You're probably best off doing them in Python since it's a simple language and has a lot of NLP libraries. One of the core classes requires Python since you have to use NLTK. Although there's a lot of programming, you aren't held to any standards like you'd be in a CS class. Your code basically just has to run. 

 

As far as after you graduate, there are many routes you can go. A lot of people go on to do NLU/NLP at the big companies like Amazon and Microsoft. A lot of people, including myself, went on to just do straight up Software Engineering. Some other people have gone the more machine learning route. Other people go on to do research.

 

Overall, it's a very diverse field and program. There's about a 50/50 mix of people coming in with linguistics vs CS. There are a lot of group assignments and people can play to their strengths. Even if you're not the strongest programmer, you can team up with someone who can handle more of the technical bulk. They're aware of the fact that people come in with different levels of experience, which is why they're pretty lenient, but they still have some minimum programming standards.

[ides-of-july]

Thanks, that's a very helpful rundown.  A question if you don't mind: Did you then go on to *study* software engineering, or just start working?

 

[Gnome Chomsky 2.0]

I started working. I did recently complete a graduate certificate in Software Architecture and Design, but this was a few years after I already started working as a software engineer.

Long story short, the UW CLMS program has a internship option instead of the thesis. I took up an internship at a startup in Seattle doing computational linguistics work. After completing the internship, I stayed with the company as a software engineer, working on the back-end of their web app. After a few years there, I got a job at a big company as a software engineer. I don't see myself going back to doing pure computational linguistics since I really like what I'm doing now.