Last night, your host attended a mixer for computational linguists from across Southern California. It was hosted by a friend with many more years of experience in the field than SC has, who performs this service to the community annually.
It was also one of the saddest sights SC has seen outside of a soup kitchen. Before, the requests for help finding a job were a trickle; last night, it was a flood. About a dozen students either recently finished with degrees, or about to be, all reported that they had no luck finding work in the field. People from two companies reported that their offices were either closing or moving. And various faculty members in attendance from three universities had little optimism to offer from their own contacts farther afield.
Quite simply, right now the demand for language technologies is almost nil. It's hard to explain why this might be the case, but it's certainly true that the public has not seen many convincing examples of why they need natural language processing. A venture capitalist that SC spoke to recently opined that the market for machine translation wasn't likely to sustain a business at even $10 million a year in revenue. Given the billions spent every year on translation services already, surely this is wrong -- but the fact that a presumably savvy investor could think such a thing suggests a disastrous failure in public relations on the part of linguists.
There are not all that many academic positions available within the field in any given year, and that situation is only likely to get worse over the next decade. In order to behave ethically towards interested students, the linguistics community has to do a better job creating excitement in the commercial world for what it does, or the value of a graduate linguistics degree will vanish. It would be cruel to keep bringing students into the field, only to leave them unable to get a job after a half-decade of specializing. But it's not clear to SC how this situation is going to be turned around.
I have to say that your experiences are not in keeping with my own. At the recent HLT/NAACL, there were more jobs being advertised than I have seen for a while, and most business-savvy people I talked to agreed that there was an increasing in the funding for small companies. I've noted in my LiveJournal that there is a lot of venture capital behind information extraction, driven by homeland security, and also by the financial information and biopharm industries.
On the other hand, when I am reviewing resumes of computational linguists who are fresh out of grad school, my first reaction is to throw most of them in the bin. This is because the L is often greater than the C. There is a lack of skill in the practice of building linguistic software: how you write robust, scalable, documented software within the context of a development team. What's worse is that there is often a lack of understanding that these issues are important, or even a outright dismissal of their importance.
Perhaps what I am saying is that there are jobs for practitioners of NLP and language engineering. I have seen very few commercially successful systems which show measurable benefit from people who can parse for food; I've seen plenty which have succeeded as a result of applying simple NLP with an attention to detail and an eye on what makes a product successful.
Posted by: David Elworthy | May 17, 2004 at 03:27 PM