Home > Uncategorized > Computational Linguistics in the Real World

Computational Linguistics in the Real World

I was recently invited to present my thoughts on how students from the University Of Ottawa Linguistics Department could benefit career-wise from the knowledge of computational linguistics. I gave an overview of computational approaches to linguistics, listed a number of areas where viable products can be created, and described three technologies that have already been successfully commercialized, or are ready for it.

Computational Linguistics in the Real World SlidePack [PDF format download]

I also provided an overview of machine translation and speech recognition/transcription as well, even though these are not reflected in the slides.

After the presentation, I asked a very interesting question (among others) that I’d like to elucidate on a bit. The question was regarding a statement I made that statistical methods in computational linguistics have become very popular, and I was asked what does this imply for the symbolic logic (rational methods).

There had always been a bit of back and forth between the empiricists and the rationalists in computational linguistics. However, with the cheap computational resources and abundance of data available nowadays, it makes a lot of sense to run some [empirical] exploratory data analysis experiments, and carry out some collocation/correlation analysis before getting really deep into the problem. This way, you can get some results within two week, rather than finding out the viability of your hypothesis after six months of intense [rationale-based] study and experimentation. If the initial results are promising, and the research problem is worthy of further study, only then should you commit yourself to studying this particular issue deeper.

  1. Numan Sheikh
    October 7th, 2009 at 04:09 | #1

    Well, as more of a theorist, i’d say that 100% evidence based on samples and statistics, doesn’t guarantee the success… although for all practical purposes things work well the way you have suggested. Its just a matter of taste probably that statistical evidence lack satisfaction

  2. October 7th, 2009 at 09:19 | #2

    Numan, your interests are not confined to theory only. I’ve seen you in actions, you’re a practical getting-things-done kind of person :-)

    As for theory, it is very valuable, and is regarded as the ‘high road’ of science. The final objective for us all is to uncover the truth, the true nature of things, and to share it amongst ourselves for the betterment of society and the world.

    If this truth can be discovered on the basis of previous theories, the results are usually quite powerful and fundamental. However, in many cases, the previous theories may have flaws within them as well, in which case, nothing is better than to visit the original data to see what reality has to say for itself.

    There is a science and an art to both methods, and I am happy that we live in a world where both are honored, respected, and efficacious.

    – Shahzad

  1. No trackbacks yet.