Textual Data in Machine Learning. Building predictive models in practice
This course covers the essential steps for building a classical SVM text classifier, neural network based classifier (CNN), and some techniques for RNN in text generation. The first session is about data preprocessing and a linguistic part. The second is dedicated to the theory and practice of building classic textual classifiers. And finally in the third session, we’ll discover opportunities and challenges given by cutting edge neural networks research.
Python3.5, virtualenv, virtualenvwrapper Jupyter notebook Scikit-learn (and satellite things like NLTK, scipy, numpy etc.) Theano/Tensorflow (optionally), Code and setup instructions provided by author are compatible with Ubuntu 15.06.
Students can use native system or virtualboxed one. It is likely that experienced MacOS user would be able to setup things there easily, but this is out of materials and course scope.
Attenders should be able to read/write simple python code.
Some experience in ML/NLP would be a plus.
Being familiar with virtualenv, virtualenvwrapper, Jupyter notebook would help.
Founder at ai-labs.org, researcher, Software Engineer and Data Scientist. Primarily work with natural language processing related projects