Thesis season, a retrospective

I wrote this in April 2016, when I was in the middle of... well, everything I talk about here. I rediscovered the draft and realized that this season is coming up for others, so I thought I would share for those who have an opportunity to avert disaster before it befalls. If you are approaching your own thesis season, know that even if you end up with a product you could not have imagined when you started, it's okay. You will finish, and your work will be worthwhile. You have no idea how many people will find it useful!

It's quite an adventure to be finishing a thesis, looking for a new place to live, and expanding a career in a new dimension. Plus making a brand-spanking-new website for yourself. Also, it's tax season, and all you self-employed people know what that means. (If you don't, thank your lucky stars and the IRS. Or just your stars.) All that said, as my quotation in Design Work intimates, that's kind of how I like to roll.

The reason I feel that this thesis deserves a blog post is that it's so absolutely not the thesis I started out thinking I would write. If you're a student about to write one of your own, I'm writing this to you, in the hope that it will spare you some of the agony I underwent before accepting that difference between what I had hoped and what became.

When I started, I submitted a proposal as required to my advisor. It bears noting that this was a brand new advisor, as my old one had suddenly gone on to fame and fortune in the corporate world, far from the ivory tower. I had discussed my ideas with my advisor emeritus, and he had gently suggested that perhaps my reach was exceeding my grasp. My new advisor told me the proposal would require four to six years, and probably several other people (not unlike Mitch Hedberg's response to a New York deli sandwich). So I pared it down, painfully lopping off really interesting and fun avenues of inquiry, until I had something we thought I could work with in the time I had (I truly never before appreciated how short a time three months is).

My education, up to this point, had been in social sciences. No technological hurdle impaired my ability to do classwork or thesis work in political science or linguistics. It turns out that when you add computation to linguistics, it can really do a number on you. While my over-all plan was working, the data produced was not in keeping with the standard for meaning imposed by the theory behind it. That is, I could get patterns out of the data, but they wouldn't really be meaningful. There were details missing that needed pre-processing steps to capture: co-reference resolution and shallow parsing.

The story of my pursuit of actionable co-reference resolution is a post unto itself. A shorter, equally rough-ending story is that of my attempt to enact shallow parsing (the computationally easier task of the two). All this really amounted to was that I couldn't make my project work as intended and write it up in the time required. My advisor advised me to write a bit of a different paper, a future-facing one about what I planned and why, rather than what I did. That way, I could at least show the thought work behind my project.

It would not be honest of me to pretend I was anything but monstrously disappointed. My thesis being the core of what I'm doing these days, it came up in a few conversations, and that's when the surprise came: a lot of people have had to do this. Obviously, the ideal is to plan, execute, and analyze an actual experiment, but a clear, well-written paper that takes the reader through all the thinking behind that experiment, whatever its results, is still valuable.

As much as I would love to revolutionize my field with what I've done (as my original advisor repeatedly told me was neither reasonable nor particularly desirable in the confines of a Masters thesis), science is a step-by-step journey through an open world. Finding out what paths not to go down and getting even general directions about where to head are useful.

I'm a little surprised that even then, I could bring myself to write down that this crushing feeling of un-success could be misleading, and the thing that prompted the feeling at all valuable. I am grudgingly impressed with Past Me. She was right: as of January 2018, I was one of the most popular authors in the digital commons that hosted my work, despite my conviction to the contrary. Try your thing. Do what you can. It's going to be okay.

#gradschool #thesis #masters #linguistics #computationallinguistics #nlp