To address this, researchers applied a technique known as "zero-shot classification," which leverages a large language model (LLM) to classify data without the need for pre-prepared training data.