0.25
0.5
0.75
1.25
1.5
1.75
2
BayesANIL - A Bayesian Model for Handling Approximate, Noisy or Incomplete Labeling in Text Classification
Published on Feb 25, 20074546 Views
Related categories
Chapter list
BayesANIL A Bayesian Model for Handling Approximate, Noisy or Incomplete Labeling in Text Classification 00:06
Outline00:25
Motivation - hurdles in supervised learning of text classifiers00:37
Related work01:29
Related work (contd.)02:52
What we propose04:00
Role of BayesANIL in text classification 05:09
The BayesANIL model : notations07:27
The BayesANIL model : notations…08:30
The BayesANIL model: Objective function09:47
The BayesANIL model: E and M Steps10:33
The Algorithm11:09
Re-estimating the empirical distribution12:42
Utilizing parameters of BayesANIL in NB14:09
Utilizing parameters of BayesANIL in SVM 14:58
Experiments and Results16:00
Experiments and Results: Supervised 17:03
Experiments and Results: Labeled-unlabeled18:23
Experiments and Results: Access to unlabeled for WebKB19:20
Experiments and Results: Access to unlabeled for 20 Newsgroups20:18
Experiments and Results: Noisy Labels21:42
Comparison with results as reported by (Bing Liu et al 2003)23:21
Experiments and Results: Notion of Support24:05
Summary25:13
Future work26:25