Lecture19-naivebayes2

This preview shows pages 1–3. Sign up to view the full content.

10/13/2011 1 CS 479, Section 1: Natural Language Processing Lecture #19: Naïve Bayes (cont.) Thanks to Dan Klein of UC Berkeley for many of the materials used in this lecture. This work is licensed under a Creative Commons Attribution Share Alike 3.0 Unported License . Announcements Reading Report #7 Nigam, Lafferty, McCallum on MaxEnt for text classification M&S 16.2 Due: today Project #2, Part 1 No pair programming you may cooperate to understand the concepts (must acknowledge) do your own work (code, experiments, report) Proper name classification with Naïve Bayes (using the categorical event model) with class conditional LM Early: Monday Due: next Wednesday (in one week) Writing matters: Use good technical writing skil s Use mathematical notation Mid term Exam Next week: Thu Sat Objectives Understand the two (three) event representations for Naïve Bayes Classification Question How to represent ܲሺݓ |ܿሻ , the “local model”? Let’s look closely at two possible candidate distributions … ' () ( | ) (| ) (' ) ( | ' ) i i i c i Pc Pw c Pc d P cP w c    Multivariate Bernoulli Model One bernoulli random variable for each word type in the vocabulary either it’s present in the doc. or it’s not Incorporates explicit negative correlations Can do feature selection: e.g., keep words with high mutual information with the class variable c v 1 v 2 v |V| . . . . . .

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
10/13/2011 2 Multivariate Bernoulli Model the puck Sports the forward kicked the ball Sports the ball Sports the President is on the ball Politics Congress resumed session
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}