You don't really need an entire LLM to do this - lightweight encoder models like BERT are great at sentiment analysis. You feed it an arbitrary string of text, and it just returns a confidence value from 0.0 to 1.0 that it matches the characteristics you're looking for.
You don't really need an entire LLM to do this - lightweight encoder models like BERT are great at sentiment analysis. You feed it an arbitrary string of text, and it just returns a confidence value from 0.0 to 1.0 that it matches the characteristics you're looking for.