Text this: Enhancing Semi-Supervised Learning With Concept Drift Detection and Self-Training: A Study on Classifier Diversity and Performance