Matt and Alistair discuss the difficulties in finding qualified annotators for labeled data in NLP research. Alistair highlights the impracticality of paying expert annotators due to cost constraints, suggesting the use of natural labels like hospital discharge diagnoses for more scalable approaches.
