3 comments

[ 2.6 ms ] story [ 18.7 ms ] thread
Hi HN,

As an effort to simplify the product research experience, we have been working on a GPT-3-powered review aggregator and analysis pipeline. For this project, we applied the pipeline to popular dental aligner products

Feedback is welcome.

Technically I am sure it's brilliant. But part of me thinks how do I trust this in the world of fake news and reviews. Your bot probably takes everything at face value.

To make sure I trust something I need to read at least 20 or so original reviews and determine if they are genuine or fake. Reputation management is a thing and an industry. Fake reviews is a business model, both in pushing up your product and burying the competition.

I hear you. Currently, the reviews are scored based on an ML model trained by Amazon reviews and their respective helpfulness rating, which helps weed out low-quality reviews.

Coming soon is a fake detection algorithm that runs a statistical analysis to spot discrepancies and contradictions among reviews from multiple sources. Additionally, more trustworthy review sources, like Reddit and Wirecutter, will be given more prominence.