The correct option is Few shot prompting.

The prompt includes labeled examples of how to classify sentiment and then asks the model to classify a new post. Providing a couple of demonstrations within the same prompt teaches the model the pattern to follow, which is exactly how this technique works. The analyst shows the model what a Positive and a Negative example look like, then requests the sentiment for a Neutral case, so the model infers the format and the decision boundaries from the given demonstrations.

Role prompting is not being used because the prompt does not set a role or an identity like instructing the model to act as a sentiment analyst. It instead relies on examples to guide behavior.

Prompt chaining is not being used because there is no sequence of dependent prompts where the output of one step becomes the input to another. This is a single prompt with embedded examples.

Zero shot prompting is not being used because the prompt does not ask for a classification without examples. It explicitly includes examples, which makes it the opposite of zero shot.

Full AWS Practitioner Certification Question