Algorithm v. 2.1 is now live
We released a new version of the algorithm which no longer requires seed accounts – it can now find communities from a single word input (e.g. “Bitcoin”, “cybersecurity” or “vegan”).
This release builds on the previous update, but it is a critical milestone. It will have a major impact on what you will be able to use hive.one for in the coming months. This post is going discuss the following consequences of this release:
Changes in the ranking
Improved ability to map more communities
Communities can have their IDs now
Naming and resolving community schisms
Changes in the ranking
You may have noticed that the ranking both in Bitcoin and Ethereum lists have changed. We have made small modifications to the algorithm that should result in increased accuracy. The new approach also enabled us to drastically increase the size of the data set, which also leads to accuracy improvements.
Mapping new communities
With this approach we can dramatically reduce the time it takes us to map a community.
Time to map a community weeks → hours
We have tested our algorithm on various types of groups and it works remarkably well across:
Cryptocurrency communities (e.g. Bitcoin, Ethereum, Dogecoin)
Communities around brands (e.g. Tesla, Apple, BMW, Roam Research)
Programming languages (e.g. Python, React, Haskel)
Professions (e.g. UX Design, Data Viz, Economists)
Interests (e.g. Cars, Space, Psychedelics)
Lifestyle (e.g. Vegan, Brazilian Jiu Jitsu, Carnivore)
Industries (e.g. Venture Capital, Biotech, FinTech)
Disciplines (e.g. Mathematics, physics, economics)
Sport teams (e.g. La Lakers, Manchester United)
Colleges (e.g. MIT, Harvard, Lambda School)
And many others we could not think of yet. We are estimating that there are millions of communities that we can potentially map. This means that we will have to find a way to prioritize through crowdsourcing. You can submit your suggestion here if there is a community you would like us to map.
Unique IDs and the new social graph
We can now give unique identifiers (IDs) to communities which we all knew existed, but we could never point to them. In other words, it is now possible to represent a community as a node in a graph.
This means that a new type of a social graph is possible. An individual user could follow (or be connected in some other way) to a community, and a community could follow a user or another community. This also opens up the way to mapping communities across multiple social media platforms.
Naming and schisms
With this new approach, naming becomes easy. The phrase we used to find the community becomes its name. Sometimes, there is more than one community that self-identifies with a given word. “Bitcoin” is a good example of that.
There is more than one community that would refer to itself as “Bitcoin” or “Bitcoiners”. They associate with different assets (BTC, BCH, BSV) and they hate each other. Resolving this is easy. The community cluster that has amassed the highest Attention score in the overall social graph gets to claim the word. In this case, it’s the community of Bitcoiners associated with BTC.
This mechanism is analogous to how blockchains resolve forks. The chain that amasses more hashpower is considered to be the continuation of the pre-split network. When Bitcoin split into BTC and BCH there was also an analogous split in the Twitter communities. BTC won the hashpower struggle and the struggle for attention on Twitter.
Community schisms like this are going to happen in the future. We think that this is a good mechanism for resolving which community should keep the preferred name.