3 ms·

Because it's a really, really hard problem to solve. The sheer scale of Amazon combined with razor thin margins and the fact that Amazon isn't even the selling

by inoop 8y ago

Because it's a really, really hard problem to solve. The sheer scale of Amazon combined with razor thin margins and the fact that Amazon isn't even the selling party on most products means that they can't just throw a bunch of manual labor at the problem. It's a similar problem to YouTube still hosting tons of copyrighted material. Fraud is just super hard to fight.

orthecreedence 8y ago
It has to be somewhere between "not enough resources to fix" and "wow, we are making a ton of money off this 'problem!'" There are so many patterns of low-hanging fruit: "I haven't used it yet but it looks great" "Seems ok" "It's useful" "The shipping was slow" On any given product with any measure of popularity, probably 10-20% of reviews consist of those formats. One-word sentences with some kind of bland/useless assertion. You're telling me the people who have massive clusters of machines and who resell tools for machine learning can't at least pick off the easiest 20% of shitty or fake reviews? Hell, give me two days, a few regexes, and a script and I'll do it myself. It's really that easy. They just don't care. That said, yes, beyond that lower-hanging fruit, the problem is insanely hard to tackle. But why not crowdsource it? People posting reviews saying "Sandy Hook was a liberal hoax" on a book about the Sandy Hook shooting could be flagged and marked for review.
- kian 8y ago
  If it's that easy to fix, and helpful, a chrome or firefox extension seems warranted. Do you have two days to spare? Let me know when you're finished, I'll definitely use it. ;-)
  - orthecreedence 8y ago
    It would be much harder to fix on the browser side, because you'd have to page through all the reviews (sometimes tens of thousands) and remove fakes, then re-adjust the star amounts. That said, it could be done...would probably take more than two days without direct access to the data though!
- TeMPOraL 8y ago
  > But why not crowdsource it? And now you have two problems :). Reviews themselves are already crowdsourcing. And any kind of crowdsourcing can and will be gamed as soon as someone can make money out of it.

by inoop 8y ago

orthecreedence 8y ago

It has to be somewhere between "not enough resources to fix" and "wow, we are making a ton of money off this 'problem!'" There are so many patterns of low-hanging fruit: "I haven't used it yet but it looks great" "Seems ok" "It's useful" "The shipping was slow" On any given product with any measure of popularity, probably 10-20% of reviews consist of those formats. One-word sentences with some kind of bland/useless assertion. You're telling me the people who have massive clusters of machines and who resell tools for machine learning can't at least pick off the easiest 20% of shitty or fake reviews? Hell, give me two days, a few regexes, and a script and I'll do it myself. It's really that easy. They just don't care. That said, yes, beyond that lower-hanging fruit, the problem is insanely hard to tackle. But why not crowdsource it? People posting reviews saying "Sandy Hook was a liberal hoax" on a book about the Sandy Hook shooting could be flagged and marked for review.

kian 8y ago
If it's that easy to fix, and helpful, a chrome or firefox extension seems warranted. Do you have two days to spare? Let me know when you're finished, I'll definitely use it. ;-)
- orthecreedence 8y ago
  It would be much harder to fix on the browser side, because you'd have to page through all the reviews (sometimes tens of thousands) and remove fakes, then re-adjust the star amounts. That said, it could be done...would probably take more than two days without direct access to the data though!
TeMPOraL 8y ago
> But why not crowdsource it? And now you have two problems :). Reviews themselves are already crowdsourcing. And any kind of crowdsourcing can and will be gamed as soon as someone can make money out of it.