Note: This post is very much my own personal experience. Yours will likely vary a lot! Please don’t feel discouraged if you don’t feel you match this - I would always recommend just going for it if you can!

Over the past year, I applied to a tonnes of schools for PhD programs, interviewed at a bunch, and got into a few. This post will be a mixture of reflections and explanations about how this process went, what I did to prepare, and how I felt during it. There’s already a number of fantastic explanations, interviews, and stories about getting PhDs in NLP and ML more broadly, and if you are thinking about or preparing to do a PhD in NLP I really recommend all these, including:

There are so many more resources out there (for example, see this list), and I definitely recommend sitting and reading as many of these as you can! With this post, I’m not aiming to target anything in particular, but just give another datapoint that people might appreciate.

T-Minus One Year

My journey to seriously starting to prepare for PhD applications begun with starting as a predoctoral young investigator at AI2. This basically gave me the chance to work on a few research projects with some great, immensely talented folks, and get lots of advice / help with applying. This is obviously something that is fairly rare, but it definitely majorly helped my chances, and made me personally much more confident in applying (since it gave me the chance to get some projects under my belt). Before the program, I had a little research experience, but no published papers, and was definitely not feeling up to applying to programs - although I know and have heard of folks succeeding in applications in similar scenarios.

Over the initial PYI year, I managed to publish one paper, and put out two others (which were under review by the time I applied). Maybe not the most productive or impactful researcher out there, but honestly the bigger thing was being exposed to the broader NLP community and gaining the confidence to talk about my research ideas and NLP research more broadly with the community at large. It turns out sometimes just being confident enough to talk about your ideas can get you some places!

T-Minus 3 Months

Around August, I started to look around and work out where I wanted to apply, and what the deadlines, professors, programs, and so on were like. I had a draft list bouncing around all year, but this is when I think I started to really take it more seriously. For every school I could think of (i.e., had heard something about, or knew about a professor there), I listed (a) the professors there and their areas of interest, (b) the application deadline, (c) any notable details about the program, and (d) general positives or negatives that I could think of (e.g., location, what I had heard from their students, so on). I tried to harden this into a partially-ordered list, segmenting the schools into ‘tiers’, and trying to pare it down so that only the places I would genuinely accept if given an offer were on the list. I also shared the list with some folks, who helped suggesting places I hadn’t heard of or considered before.

The final list was still long (I ended up applying to ~13 schools), but going through the process helped focus my efforts a bit.

T-minus 1.5 Months

Around the start of October, I started to draft up my statement of purpose (SoP) and ask folks to write letters of recommendation for me. After some time working on the SoP by myself, I shared it around with a few folks and got back some really good, clear feedback. In the end, I probably rewrote the SoP almost from scratch 3-4 times in order to really shake it up and keep it tight. I definitely recommend reading the SoPs on CS SoP for inspiration, and sharing your drafts around (this can be scary, but the feedback is definitely worthwhile!). It’s cliche to say, but try to stay positive and not compare yourself to too many of the (legitimately insanely good) candidates out there - just do what you can, without losing your mind over it.

T-Minus 1-2 weeks

Right before the first deadlines started approaching in late november, I basically started creating accounts in the grad application systems and checking to see what documents I would need. Something worth noting is that every school often has unique constraints or things to do for its application: most schools constrain the SoP in some way (I saw 2 pages, 1 page, 1000 words, single-spacing, double-spacing, pasting raw text only with no formatting). Additionally, some schools require a personal statement (talking more about your own experience and goals), and some require random short answer questions (e.g., write about a time you had to teach).

I ended up spending most of the week before and of thanksgiving writing the first batch of these, targeting the earlier applications first (warning: European deadlines are much earlier than the American deadlines, and you might have to be ready in october or early november). This first hump was the toughest, but once I was through it I had a bunch of variants of personal statements tailored for a bunch of different universities, with the most useful two variants were the two-page and 1000 word version, since shrinking two pages down to 1000 words requires a non-trivial amount of cutting.

Much of this time was also spent inputting my details into the varying application systems, which was mind-numbingly boring and took ages at times. You should really do this because often you can log in and send out recommendation letter requests once you have created an account, which gives more time for your letter-writers to work out the system and send in their letters.

The Thick of It

Over December, I attended EMNLP and during this continued the application process. At this point, I was just trying to stay a bit ahead of the deadlines while also attending the conference. I mostly had the parts I needed, as noted above, but it still felt like a frantic rush, as I often was submitting 2-3 applications a day and still had to write the final paragraph of the personal statement where I tried to make connections between my work / interests and the professors I was applying to explicit (some people recommending not doing this, I’m not sure what is best).

Finishing up the Applications

I finished up applications by December 15 while in Australia (some schools have later deadlines, but I did not go for them). Unfortunately, there was a paper deadline 5 days later, so I had to turn my attention to paper-writing, but after that I just took a massive break and basically didn’t touch my laptop at all for two weeks.

T-plus(?) 2 weeks

Starting in very late december (around the 28th), I started to get interview requests (although this was very early). This then increased a bit in the first week of january, and up until around the end of January (~the 25th), I started to get emails asking for interviews on a regular basis. These were actually pretty spread out - I think I almost always had at least one interview lined up at any given point from late December to late January. These were a bit nerve-wracking, of course, but were all ultimately enjoyable and not too difficult. It turns out that talking about research and NLP is fun, actually! I did prepare for these a little, collating a list of common questions and preparing answers, and I sort of refined this and worked on it over time. For each professor, I looked up their background and recent work, and tried to think about questions specific to that school or professor that would be useful.

Here’s a (abbreviated and paraphrased) list of questions I prepared for and got from multiple people. I don’t think any of these should be too surprising!

  • Why do a PhD in NLP?
  • (relatedly) Why a PhD and not industry work or research?
  • What is your ideal advisor / research group / grad school experience?
  • Tell me about a recent project you’re especially proud of.
  • What are your broader research interests?
  • What is some recent work you were interested in? Can you explain a bit about it?
  • What do you think is the future of NLP?

For the large part, questions were not too technical, but definitely be prepared to talk about your own work and maybe 1-2 recent papers in detail, to be safe. Although - every professor has their own style, so it’s hard to give one-size-fits-all advice.

Also - I did not look at gradcafe or similar through this process, and to be honest didn’t ever feel the urge to. Everything runs at its own pace, and at this point you just have to go with the flow!

T-plus 1-2 months

I more or less was done with interviews by the end of January, and I started getting offers in early February. It’s worth noting timelines here seem very, very messy - I know folks that got offers from the same school at different times, and even right up until April some folks I know were getting offers. Every school has its own pace, and I think the best strategy once interviews are done is to just try and not think much about it and don’t expect super-fast turnarounds.

T-plus ~4 months

I took a nice long break vacationing around Europe to enjoy the period before starting PhD, and I definitely recommend taking some time off if you can afford it!

Conclusion

This was a very rough timeline with some thoughts. Don’t worry if you are ready sooner or later than the dates here, and feel free to move at your own pace. But hopefully it provides a little insight into what (I think) a semi-reasonable timelines looks like. It’s also worth noting that applying for things like fellowships, NSF often take place earlier than admissions, which can be another chance to get stuff ready early.