Concept labeling: building text classifiers with minimal supervision

摘要: The rapid construction of supervised text classification models is becoming a pervasive need across many modern applications. To reduce human-labeling bottlenecks, new statistical paradigms (e.g., active, semi-supervised, transfer and multi-task learning) have been vigorously pursued in recent literature with varying degrees empirical success. Concurrently, the emergence Web 2.0 platforms last decade has enabled world-wide, collaborative human effort to construct massive ontology concepts very rich, detailed accurate descriptions. In this paper we propose framework extract supervisory information from such ontologies complement it shift direct labeling examples domain interest much more efficient identification concept-class associations. Through studies on categorization problems using Wikipedia ontology, show that allows high-quality be immediately induced at virtually no cost.