Text Classification
Overview
Text classification maps documents, sentences, or messages to categories, from broad topics and sentiment to specialized labels such as legal outcomes, medical findings, and safety risks. Common approaches range from TF-IDF with linear classifiers to fine-tuned transformer encoders, large language model prompting, retrieval, and methods that use label hierarchies, linguistic features, or multimodal evidence. Current work also focuses on learning with few labels, class imbalance, multilingual data, and efficient systems that combine automated predictions with human review. A central challenge is whether strong benchmark scores hold up on new domains, languages, long-tail classes, and ambiguous examples, while keeping costs and errors manageable.
Latest papers
226 papers31 papers in the last four weeks, up 63% on the four weeks before. 0.2% of all new papers.