Active Learning Strategies for Efficient Text Classification
A.R. Desai · 2025
The exponential growth of textual data in electronic form has made automated text classification essential for organizing and retrieving relevant information. Traditional supervised learning approaches require large amounts of labeled data, which is often costly and time-consuming to annotate. Active learning offers a promising solution by selectively querying the most informative examples for labeling, thereby improving model performance with fewer labeled instances. This paper explores various active learning strategies, including uncertainty sampling, query-by-committee, and representativeness-based models. Experiments are conducted on real-world datasets, such as Taobao reviews and e-commerce customer queries, using classifiers like Na¨ıve Bayes, Support Vector Machines, and logistic regression. The results demonstrate the effectiveness of active learning in enhancing classification accuracy while minimizing labeling effort.