Classifiers take some sort of input, text for us, and assign it to one or more categories. This very simple capability forms the core of modern NLP techniques like sentiment analysis, named entity detection and information extraction. This class will cover both heuristic and machine learning based classifiers from the perspective of a programmer building such systems.
LingPipe is an open source API designed to build NLP systems. Breck Baldwin instructs the class.
The class covers rarely addressed issues around creation, training and maintaining classifiers with a focus on modern learning techniques like logistic regression. This is a practitioner level course with minimal theory appropriate for proficient developers. The class will be taught with real time coding and those that wish can follow along on their own laptop.
The course outline is:
- Introduction to classifiers and information they provide ranging from first best classifiers to conditional probabilities.
- Creation of gold standard data with tweets
- Classifier evaluation a.k.a. the dreaded evaluation harness
- Heuristic classifiers
- Machine learning classifiers
- Feature tuning for machine learning
- Common problems encoded into a classifier framework
The class is 6:30pm to 9:30pm Tuesday, October 25 at 181 N 11th St, Suite 401, Brooklyn NY 11211. The location is 4 blocks from the Bedford L stop and is a 4 floor walkup. There will be pizza served as a break and those that wish to have a beer afterwards will meet at the famed Mugs Ale House a block away.
The cost is $100 with a class limit of 15. Registration closes October 23. Payment via PayPal to paypal@alias-i.com.