A developer wrote the following prompt to classify the sentiment of a user comment: 'What do you think about this? Input: The new user interface is confusing and slow.' Analyze this prompt and explain two specific weaknesses that make it ineffective for a zero-shot classification task where the desired outputs are 'positive', 'negative', or 'neutral'.

Google

A zero-shot prompt applies a task instruction to a new input without any preceding examples. To classify the sentiment of a positive text, the prompt could be structured like this:

'Assume that the polarity of a text is a label chosen from {positive, negative, neutral}. Identify the polarity of the input.

Input: The weather here is wonderful.'

Example of a Zero-Shot Prompt for Polarity Classification (Positive Sentiment)

A developer needs to determine if a customer review is 'positive', 'negative', or 'neutral' without providing any examples in the prompt itself. Which of the following prompts is best structured to accomplish this task reliably?

Critique of a Zero-Shot Prompt for Sentiment Classification

You are tasked with classifying the sentiment of the following customer review: 'The user interface is confusing and difficult to navigate.' Construct a complete, self-contained prompt that instructs a language model to perform this classification without providing any examples. Your prompt must explicitly state that the output label should be chosen from the set {positive, negative, neutral}.

Learn Before

Related