Open herongrove opened 7 years ago
@danebell: This is the paper I mentioned: http://nlp.stanford.edu/~sidaw/doku/doku.php?id=docclass:overview I found the "very brief overview" and the slides to be easier to understand than the paper itself.
Also, we should think about how to add recency info to this model.
Instead of raw counts, experiment with using log-odds features (as in Bayes, but in the SVM)