It is a great question, and one that we are still working on doing effectively. We've essentially built our own product taxonomy that started from Google's taxonomy and then branched off / enhanced from there based on our use case. You can build NER / NEL models from there to help identify which products go in which categories. Hope this helps.
Awesome. Beni is definitely going to be a great product.
If I understand correctly, after the classification, the number of pictures that need to be compared in each bucket should not be too huge. But I am still curious, could you introduce a little about your algorithm related to the similarity of pictures? In this regard, any existing AI tools can help?
Thank you, I am glad you like it! You can actually go download it for free on the Chrome Web Store (desktop only) or the Apple App Store (available as an iOS or macOS extension) if you want to test it out.
For the pictures, we vectorize all of our images using an embedding model and then we do a similarity search through the vectors to find similar pictures. Hope this helps!
Celine, your work is so great! Thanks Bryant for sharing the post with me.
If possible, I would like to ask Celine, with so many eCommerce sites and so many product categories, how do you organize them reasonably?
It is a great question, and one that we are still working on doing effectively. We've essentially built our own product taxonomy that started from Google's taxonomy and then branched off / enhanced from there based on our use case. You can build NER / NEL models from there to help identify which products go in which categories. Hope this helps.
I feel that Beni has integrated many fields of AI. Looking forward to your next series of introductions related to NLP processing!
Best wish for you and Beni!
Awesome. Beni is definitely going to be a great product.
If I understand correctly, after the classification, the number of pictures that need to be compared in each bucket should not be too huge. But I am still curious, could you introduce a little about your algorithm related to the similarity of pictures? In this regard, any existing AI tools can help?
Thank you, I am glad you like it! You can actually go download it for free on the Chrome Web Store (desktop only) or the Apple App Store (available as an iOS or macOS extension) if you want to test it out.
For the pictures, we vectorize all of our images using an embedding model and then we do a similarity search through the vectors to find similar pictures. Hope this helps!
Tysm! Celine. I'm overjoyed by your response.
Thanks for this breakdown! The first visual was really helpful in understanding the relevant site info.