A few weeks back I gave a talk about Representation Learning and Sequential Models for Text Processing at PyCon Finland 2016. The talk was based on the work I did for my Master’s thesis that I recently finished. My thesis was an explorative project in collaboration with Oikotie Työpaikat, a popular job and recruitment platform in Finland which is part of Sanoma. As a result of exploring and reframing the problem scope iteratively I focused on classifying text in job listings. I made use of methods from recent advances in learned Distributed Representations and in sequential language modeling with Recurrent Neural Networks. In case you’d like to read more, my thesis can be found on GitHub: https://github.com/cle-ment/ma-thesis-tex (it’s written in LaTeX, check the pdf for the print version).

At PyCon Finland I gave an overview of these topics and I showed a few examples of how such methods can be implemented and experimented with in Python. A recording of the full talk is available on YouTube thanks to the awesome crew organizing the conference:

The slides can be found on GitHub: https://github.com/cle-ment/pycon-finland-2016-talk. This was the first time I made a full slide deck using Jupyter Notebook with the RISE extension, and it works great especially since code can be executed live on the slides. And of course you can even edit slides on the go which apparently caught someone’s attention on Twitter (who also took a beautiful picture :D):

All in all the conference was really interesting with talks from unit testing using the scipy stack to functional programming in Python and I recommend checking it out.