Meta learning with language models: Challenges and opportunities in the classification of imbalanced text