3 ms·
Shopping on Amazon has become terrible and I've stopped doing it unless it's the only option (which is very rare). There is seemingly just no quality control at
by exogen 10y ago
Shopping on Amazon has become terrible and I've stopped doing it unless it's the only option (which is very rare). There is seemingly just no quality control at all in their catalog; it's like shopping in a flea market or junkyard and they need to take responsibility for it.
Two screenshots from last year, when I was trying to shop for electrical receptical covers:
https://pbs.twimg.com/media/CTQroCDU8AAw5U7.png https://pbs.twimg.com/media/CTQroCDU8AAw5U7.png
https://pbs.twimg.com/media/CTQqWPnVAAArZM1.png https://pbs.twimg.com/media/CTQqWPnVAAArZM1.png
The first is categorized under "Tools & Home Improvement > Power & Hand Tools > Hand Tools > Pliers > Snap-Ring Pliers".
The second is categorized under "Musical Instruments > Drums & Percussion > Folk & World Hand Drums > Timbales".
How does that even happen? Can't they use some machine learning to detect that this is so obviously wrong?
- majewsky 10y agoMachine learning wouldn't help there, because it will learn from the existing categorizations. Garbage in, garbage out.
- exogen 10y ago(1) Why would they train on the existing categorizations? You train on a smaller, known-good dataset. (2) Outlier detection is a thing, even if they included the bad data. It would be trivial to detect that a small percentage of things categorized as "musical instruments" have extremely dissimilar facets/descriptions/images to all other things in that category, and very similar to things in a completely unrelated category.