Paper: Multi-label Text Categorization with Model Combination based on F1-score Maximization

ACL ID I08-2116
Title Multi-label Text Categorization with Model Combination based on F1-score Maximization
Venue International Joint Conference on Natural Language Processing
Session Main Conference
Year 2008
Authors

Text categorization is a fundamental task in natural language processing, and is gener- ally defined as a multi-label categorization problem, where each text document is as- signed to one or more categories. We fo- cus on providing good statistical classifiers with a generalization ability for multi-label categorization and present a classifier de- sign method based on model combination and F 1 -score maximization. In our formu- lation, we first design multiple models for binary classification per category. Then, we combine these models to maximize the F 1 -score of a training dataset. Our experi- mental results confirmed that our proposed method was useful especially for datasets where there were many combinations of cat- egory labels.