llm-prompts.md

Prompts used with LLMs

{text} is classification sample text and {target} is a stance target.

System command

System command used with gpt-3.5-turbo:

You are a stance classifier

Prompt P3

From reading the tweet below, which of the options is most likely to be true about the tweeter's stance or outlook (explicit or implied) towards the target: {target}. Is it in favor, against or none?
Tweet: {text}
Answer with only one of 3 words: favor, against, none

Prompt P4

What position (explicit or implied) does the paragraph bellow take on the topic of: {target}? 
{text}
Answer with one of 3 words: favor, against or neither}

Prompt P5

An arbitrary user of social media webpage has posted the following:
{text}
You know nothing else about this user, but you should make assumptions regarding what he has posted. Make a best guess about an outlook of this user towards: {target}. Answer none if you aren't sure.
Answer with only one of 3 words: favor, against or none