An empirical study of smoothing techniques for language modeling
ACL '96 Proceedings of the 34th annual meeting on Association for Computational Linguistics
The class imbalance problem: A systematic study
Intelligent Data Analysis
SMOTE: synthetic minority over-sampling technique
Journal of Artificial Intelligence Research
Improving naive Bayes text classifier using smoothing methods
ECIR'07 Proceedings of the 29th European conference on IR research
Naive bayes for text classification with unbalanced classes
PKDD'06 Proceedings of the 10th European conference on Principle and Practice of Knowledge Discovery in Databases
| Hi-index | 0.00 |
Multinomial naïve Bayes is a popular classifier used for a wide variety of applications. When applied to text classification, this classifier requires some form of smoothing when estimating parameters. Typically, Laplace smoothing is used, and researchers have proposed several other successful forms of smoothing. In this paper, we show that common preprocessing techniques for text categorization have detrimental effects when using several of these well-known smoothing methods. We also introduce a new form of smoothing for which these detrimental effects are less severe: ROSE smoothing, which can be derived from methods for cost-sensitive learning and imbalanced datasets. We show empirically on text data that ROSE smoothing performs well compared to known methods of smoothing, and is the only method tested that performs well regardless of the type of text preprocessing used. It is particularly effective compared to existing methods when the data is imbalanced.