Tuesday, February 8, 2011

Linguist List FAIL

I've been kicked around a few NLP blocks in my time so I've developed a sixth sense about what employers are looking for when they post job announcements. When I read this one from Intelius on The Linguist List today, my reaction was clear, concise, and unconditional: This is NOT for linguists.

This posting says engineers only to me! There's nothing wrong with that, but why use the Linguist Lists' job postings board with a job that no actual linguist will be considered for? My reaction is based on what I consider to be engineering dog-whistles that are designed to encourage the "right" people to apply (i.e., engineers) and the wrong people to go away (i.e., linguists).

A quick breakdown of their rhetorical dog-whistles:
  • The Data Research Group is a team of scientists at Intelius... Much as I would like linguists to be considered scientists, the truth is, in the "real world" of job announcements, they are not. This is a red flag.
  • Team members have published papers in top research conferences...Ah hah, not "conferences" per se, but "research conferences". This means ACL.
  • Mentors will include Dr. Vitor R. Carvalho and Dr. Andrew Borthwick (diss PDF)... NOT linguists.
  • Required Skills: Strong hands-on skills in Java and/or Python... i.e., we assume you lay awake at night worrying about arrays and functions, not unnaccusative marking and tone sandhi
  • Required Skills: Self-motivated, creative, and independent researching skills ... we will teach you nothing. You are on your own. Your teachers are gone. What can you give us?
FYI: Recently, bulbul has quite rightly taken me to task for being a tad hypocritical in arguing two seemingly contradictory points: (1) that 21st Century linguists should study math and (2) that the time consuming effort of learning computational tools is a deterrent to being a linguist. I can imagine this post as falling victim to that same complaint. My pre-defense is that I believe there is a skill set distinct to linguists that is valuable and worthy of investment by NLP capitalists that has been largely ignored.

Engineers alone will not solve the critical language issues necessary to create the great products of the next generation of NLP tools. I believe in team building where linguists and engineers work together as equals 


bulbul said...

Hehe :) To clarify a little: I don't think you were being hypocritical - the position you took is understandable and I hear similar comments all the time. The problem is that sed, awk, Perl and Python have become the tools of our trade along with dictionaries and grammar books. Oftentimes better tools will be made available and thank God for them, but I think it is very important that every linguist knows the basic stuff. And this applies to math, too - regexps and ANOVA are both indispensable, no matter how much one may dislike either (or both :).

This is NOT for linguists
Oh abso-fracking-lutely. I get so fed up with people (especially those with some vested interest, coughacademiacough) touting the great number of job opportunities for linguists when in fact most of them are engineering jobs. Rule of thumb: if it says 'Java required', it's not a job for a linguist. Java is a very specific skill, one that takes years and years to acquire. I know a bunch of people with the kind of Java skill required for NLP jobs - hell, I was sitting right there next to them (reading up on English syntax and Arabic dialectology) while they acquired their skills and experience. So I know for a fact that in most cases, you either have the required linguistic knowledge or you have strong hands-on Java skills. To have both, especially these days, would require almost superhuman abilities. Or zero sleep between the ages of 18 through 28.

Chris said...

Some folks on Twitter have suggested that phoneticians are still employable in industry and I suspect they're right. Also, there are some linguists who are also programmers so they have a dual skill set.

CoffeeTeaLinguistics said...

Chris, I've heard that [about phoneticians] as well from professors, and seen how linguists who only did theory are having trouble on the job market. There's a certain logic to it in that if you didn't do particularly empirical research, you really don't end up with the same skill set at handling data, analysis, stats, evaluation, and reporting.

Phoneticians have to handle lots of data, and the data is stuff that can be of interest in industry (although you would know much more about this than me I think), compared to something like modeling a universal representation for noun phrases

Chris said...

Yes, data analysis is key, but also because the market for speech technologies is more mature/profitable. Psycholinguists are good at data analysis but their skill set has not been monetized. Same with field linguists.

Brandon C. Loudermilk said...

Chris - you just used the word "monetize"... therefore, you automatically lose a couple points in my book ;) Frankly, I have no clue what skill-set linguists have - there are way-too-many specializations to make any generalizations.

Chris said...

I'm saving "synergy" for a special occasion, hehe.

toomim said...

I'm going to say this bluntly. You are fucking lazy. You could learn how to code enough java in a month. It is way worth your time.

Chris said...

Haha, guilty for sure. But, for the record, I have in the past "learned enough" to be a working NLP professional (I actually contributed C++ code to a profitable, commercial NE system). I agree anyone can learn a lot in a month (though I would recommend Python and the NLTK).

However, it would be misleading to suggest that ALL you had to do was spend one month learning to program and then poof you're good. It's not like riding a bike. It's the opposite. You have to use it or lose it.

But more to the point, few companies would hire you if your resume said Masters in linguistics and one month learning Java. No way. Unemployable in the tech market.

bulbul said...


on the off chance you are not a troll and/or a stupid kid:
The ads do not ask for "just enough Java to get by", they demand "strong hands on experience." And that kind of experience is something people can only acquire in years of working on actual real-life projects, not from reading Beginning Programming with Java For Dummies. And you can't learn enough of anything in one month to get by, especially not a language. Care to prove me wrong? Start learning Czech now, there'll be a test on 3/20.

Putting the Linguistics into Kaggle Competitions

In the spirit of Dr. Emily Bender’s NAACL blog post Putting the Linguistics in Computational Linguistics , I want to apply some of her thou...