Local explainability as emerging functionality?

Thanks guys for the ChatGPT prompting course, very useful and inspiring.
For the lamp review in the Inferring lesson I’ve updated one of the prompts in the following prompt:

What is the sentiment of the following product review, 
which is delimited with triple backticks?

 

Give your answer as a single word, either "positive" \
or "negative".
Also capture at least 6 combination of words motivating that sentiment, 
expressing them in JSON format using key 
Textual_Reason_1 ... Textual_Reason_6
Finally, classify each of such reason and in the same order
to one of the following customer-centered categories:
- Value for money
- Fast delivery
- Useful product feature
- I feel understood by the company
- Easy of use
- I feel valued by the company
representing each of them in the same JSON structure
using keys Reason_1 ... Reason_6.

obtaining this very nice completion reporting both the verbatim combinations of words of the review motivating the sentiment and also their classification based on listed categories:

Sentiment: positive
{
"Textual_Reason_1": "additional storage",
"Reason_1": "Useful product feature",
"Textual_Reason_2": "not too high of a price point",
"Reason_2": "Value for money",
"Textual_Reason_3": "Got it fast",
"Reason_3": "Fast delivery",
"Textual_Reason_4": "company happily sent over a new one",
"Reason_4": "I feel understood by the company",
"Textual_Reason_5": "easy to put together",
"Reason_5": "Easy of use",
"Textual_Reason_6": "great company that cares about their customers and products",
"Reason_6": "I feel valued by the company"
}

Very impressive, I must say.

Hi @giggio,
Thank you for your contribution to the community

1 Like