Wrong sentiment analysis results from text-curie-001

API
1

I am using the text classification to run sentiment analysis on Tweets, but we keep getting wrong sentiment analysis, what can we do to improve the result, to give us exactly what we need. The question was “Using one word, which of Plutchik’s emotional categories does this statement express?”

Here is the prompt below.

const completion = await openai.createCompletion({
model: “text-curie-001”,
prompt: Using one word, which of Plutchik’s emotional categories does this statement express?.\n\nTweet: \"${data[i]._source.tweetText}\"\nSentiment:,
});

Sample wrong results:

Tweet: - Your Listening to Hank Mobley - Avila And Tequila - 1998 - Remaster on Detroit’s Premier Jazz Station, WJZZ Detroit
Sentiment result - Positivity. The person is listening to Hank Mobley, which is a

Tweet: - RT @trillavnilla: Idk man every time Sunday rolls around I just wanna be double fisting some tequila pineapples

Sentiment result - Mixed

Tweet: - @Sabrina420 At least it’s decent tequila and I would never judge you.

Sentiment result - Healing