Adaptive Two-View Online Learning for Math Topic Classification
Abstract
Text categorization has been a popular research topic for years and has become more or less a practical technology. However, there exists little research on math topic classification. Math documents contain both textual data and math expressions. The text and math can be considered as two related but different views of a math document. The goal of online math topic classification is to automatically categorize a math document containing both mathematical expressions and textual content into an appropriate topic without the need for periodically retraining the classifier. To achieve this, it is essential to have a two-view online classification algorithm, which deals with the textual data view and the math expression view at the same time. In this paper, we propose a novel adaptive two-view online math document classifier based on the Passive Aggressive (PA) algorithm. The proposed approach is evaluated on real world math questions and answers from the Math Overflow question answering system. Compared to the baseline PA algorithm, our method’s overall F-measure is improved by up to 3%. The improvement of our algorithm over the plain math expression view is almost 6%.
Cite
Text
Nguyen et al. "Adaptive Two-View Online Learning for Math Topic Classification." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2012. doi:10.1007/978-3-642-33460-3_56Markdown
[Nguyen et al. "Adaptive Two-View Online Learning for Math Topic Classification." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2012.](https://mlanthology.org/ecmlpkdd/2012/nguyen2012ecmlpkdd-adaptive/) doi:10.1007/978-3-642-33460-3_56BibTeX
@inproceedings{nguyen2012ecmlpkdd-adaptive,
title = {{Adaptive Two-View Online Learning for Math Topic Classification}},
author = {Nguyen, Tam T. and Chang, Kuiyu and Hui, Siu Cheung},
booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
year = {2012},
pages = {794-809},
doi = {10.1007/978-3-642-33460-3_56},
url = {https://mlanthology.org/ecmlpkdd/2012/nguyen2012ecmlpkdd-adaptive/}
}