Creative AIML Datasets

I just read about how, to train AIML to do sentiment analysis, some projects use Amazon reviews for data.

See, every Amazon review has to be accompanied by a star rating. if the star rating is poor, then the accompanying text is likely to be negative. And vice versa.

Blew my mind.

It's totally self-contained. No need to hire humans in Kenya or India to rate a piece of text to determine and label its sentiment.

I thought this was a highly creative way to approach the "data problem".

I want to learn more about such creative solutions to the "data problem".

Please share what you know about such creative solutions (or your ideas on creative solutions to the "data problem") on this page so we can have a resource to reference.

Thank you.

1 comment

[ 3.4 ms ] story [ 10.5 ms ] thread

bigyabai 1y ago ↗

FWIW Google Bert has been doing sentiment analysis with LLMs using text classifiers for more than 7 years at this point: https://github.com/google-research/bert

It's definitely a neat solution, but a pretty old one at this point. Those AWS Mechanical Turk employees lost their job a half decade ago, chances are.