Does naive bayes classifier perform text annotation (sequence tagging)? If yes, i need a tutorial please. What do you think about MALLET for naive baies?
Asked
Active
Viewed 124 times
1 Answers
1
Naive Bayes is a form of classifier that makes a prediction about one variable, for example a label for a document. In a sequence tagging problem, you are making predictions about a sequence of variables: one for each token.
You can do this by treating each token as its own independent classification problem, or you can use a model that makes predictions for the whole sequence jointly, with the decision for one token affecting the decision for neighboring tokens.
The sequence equivalent for Naive Bayes is a Hidden Markov Model. An equivalent classifier/sequence-tagger pair is logistic regression and conditional random fields (CRFs). Mallet implements all of these, as do many other systems.

David Mimno
- 1,836
- 7
- 7