Decentralising Knowledge Graph

Republished from https://medium.com/@npd.blockchain/decentralizing-knowledge-graph-5641775291c1

This article assumes that the reader has some knowledge about “linked data”. If not, you can have a quick two minutes read about it here.

Linked data favoured building a knowledge graph, as the semantics can be inferred from the data gathered. For instance, the knowledge graph built by Google, is shown on the right hand side of the search results. If you search for ‘Star wars’, it shows the results as shown in the image on the left.

This is possible with the growth of linked data. Star wars is the movie and the schema for “Movie” defines the relevant information required to describe a movie, like the director, actor, production company and so on..

Now try clicking on the cast link in the knowledge graph, you will notice a complete different set of information which will be defined in the schema for “Person”.

Why to decentralize the knowledge graph?

Knowledge graph is a powerful information and will be more powerful when the unstructured data starts converting to linked data. This information cannot reside with one organization or a with a limited few. The web3.0 (alias open web), shall be completely open, right from the schema definition till building the open graph.

With a lot of decentralized system making the news, it time we start thinking of decentralizing the knowledge graph. Though there are already a few blockchain protocols trying to bring a change, it will be great if they can follow a protocol for linked data. Eventually, building the pure open web.

Building a Pure Open Web

To enable a pure open web,

-- Anyone shall define or update a semantic schema on a decentralized network,

-- The data contributors can post the link to their data and the mining nodes can validate the data (just like verifying a transaction for double spend), to reduce the resource wastage.

-- The validated data can then be parsed and update to the knowledge graph, for which the nodes can be rewarded

Once these are implemented, then it makes sense to think making the linked data interoperable among the blockchains, where a consensus shall be reached from multiple blockchain protocols. Also, each blockchain shall provide a way to link their graph to third party, or if they do not have provision to build a graph, they may agree to provide the data to third parties who can maintain a knowledge graph. The data is public anyways.

Though this is just a concept at this point in time, there may be a few complication that may pop up in the implementation. Like, the number of nodes that the graph can support. One have to decide whether to traverse the graph or index the nodes for easy querying. The choice of RDF (Resource Description Framework) vs LPG (Labeled Property Graph) will also play a crucial role in the implementation.

Does the current graph databases technologies capable of handling such big dataset? Or will it be better to go with a hybrid data model or any other database technology will serve it better? We would like to hear more from you.

Contributors: Vishnu Govind / Vaideeswaran Sethuraman

Join our conversations on telegram