mleku on Nostr: the onboarding problem... need to have some decent recommendations or a mechanism to ...
the onboarding problem... need to have some decent recommendations or a mechanism to figure out how to generate a good one out of a set of known trusted users
i'm working on a project for my paid gig where it analyses user profile data to make recommendations of possible good follows for a person. one of the things it does is compare the text of posts made by users to create a kind of distance metric of similarity.
i'm just reading about TLSH and Nilsimsa proximity hashes, used in anti-spam systems. probably, using such a thing, you could generate a 2d map of users by their proximity to each other using tensors and then scattered across that you could pick out some samples on several different points in the map and then from there the user could explore the connections from the ones they like
this seems to me like something nobody has done yet
Published at
2025-06-16 06:38:01Event JSON
{
"id": "5fd7be5f0cde2a44f6294c184d89b960c8066344b2d9efe36d844938e2877918",
"pubkey": "4c800257a588a82849d049817c2bdaad984b25a45ad9f6dad66e47d3b47e3b2f",
"created_at": 1750055881,
"kind": 1,
"tags": [
[
"e",
"76ba9a762edabda4fb1e7135a1148301a5000cabce2a5aec85a52f7c9af64b6e",
"",
"root",
"8125b911ed0e94dbe3008a0be48cfe5cd0c0b05923cfff917ae7e87da8400883"
],
[
"e",
"9e87209315fd5f78d84f0f094070ce95016f2f7af66e228855c22cb27351b217",
"wss://nostr.mom/",
"reply",
"8125b911ed0e94dbe3008a0be48cfe5cd0c0b05923cfff917ae7e87da8400883"
],
[
"p",
"ee6ea13ab9fe5c4a68eaf9b1a34fe014a66b40117c50ee2a614f4cda959b6e74"
],
[
"p",
"8125b911ed0e94dbe3008a0be48cfe5cd0c0b05923cfff917ae7e87da8400883"
]
],
"content": "the onboarding problem... need to have some decent recommendations or a mechanism to figure out how to generate a good one out of a set of known trusted users\n\ni'm working on a project for my paid gig where it analyses user profile data to make recommendations of possible good follows for a person. one of the things it does is compare the text of posts made by users to create a kind of distance metric of similarity.\n\ni'm just reading about TLSH and Nilsimsa proximity hashes, used in anti-spam systems. probably, using such a thing, you could generate a 2d map of users by their proximity to each other using tensors and then scattered across that you could pick out some samples on several different points in the map and then from there the user could explore the connections from the ones they like\n\nthis seems to me like something nobody has done yet",
"sig": "26237e0173e37efde3ab89a91416b25f5339d79beb470e01f0c8a06f360c09c678a97fbde7b55229e17adee72babaa91ce6e267951aa93acc75dc6e774dd101c"
}
