Nostr Web Client

Hey, David. I'd like to try this. Where can I learn about the GrapeRank algorithm and the variables being used to determine ranking?

david 1y ago

Sorry Alison I just saw this.

You can try it out here:

https://grapevine-brainstorm.vercel.app

Although I have to apologize in advance if it’s buggy and freezes up for you. Been having some problems with neo4j settings. I’m thinking I may need to hire a neo4j expert to speed things along.

Right now the only variables are follows and mutes. Although in principle, the process of “interpretation” allows you to use any sources of data you want. Someone just needs to write a script to translate each new data source — zaps, for instance — into a format ready for digestion by the GrapeRank algo. That “interpretation” script assigns a meaning to each piece of data. In the current implementation, a mute is interpreted to mean “bot” and a follow is interpreted to mean “not a bot” (score 0 and 1, respectively). Interpretation also assigns a weight, a degree of confidence, to each piece of data. Example: Maybe a zap also means not a bot but you think it should carry more weight than a follow.

Reply to this note

Please Login to reply.

Discussion

Alison Avery 1y ago

Gotcha. Okay, I'll try it this coming week.

I can see the challenge of interpreting meaning out of each one of the variables. For example, the mutes. In my case, sometimes there are people with great information or content whom I choose to follow, but they post so often that I also mute them periodically so my feed isn't overwhelmed. Likewise, I periodically go back and unmute them to check in with what's going on. I want to keep following them, and they're not a bot, I just don't need a post every 2 or fewer hours hours of every day.

Just an example there of where a mute wouldn't necessarily represent a bot. But I guess in most cases it is more likely to represent a bot. There's probably a better way for me to achieve the same effect on my feed without muting (maybe it's creating separate feeds and following to separate feeds), I just haven't figured out how to do it yet.

And thanks for the bugginess warning on the upfront. I can give you feedback about what I encounter as I make my way through, if it helps.

david 1y ago

Yes please - feedback is welcome!

Part of my philosophy is that one’s interpretation of data generated by someone else can never be perfect. A mute, as you point out, can mean things other than bot. So let’s say that some day people start using reports (NIP 56), labels (NIP 32), or some other method to flag bots that is less ambiguous and more specific. It will be relatively straightforward to feed that new source of data into GrapeRank, give it more weight than mutes in calculation of the “not-a-bot” score, and improve the ability of your grapevine to differentiate bots from not-bots. This ability to consume data and turn it into useful scores will in turn create a demand for greater variety of methods to generate greater variety of types of data.

I hope that makes sense!

Alison Avery 1y ago

It does, indeed. And thanks so much for taking the time and thoughtfulness to provide the explanation. Very useful.