In spite of the widespread use of knowledge graphs, defining the right structure and populating it with data is still a challenge. As a result, enterprises fail to maintain the knowledge graphs after the initial burst of enthusiasm. Consequently, usage becomes minimal, and the knowledge graphs virtually cease existing, leading to a problem described as the ontology problem. This article offers two methods to solve the ontology problem by approaching it in a distributed manner.
The first method is to keep the ontology simple and domain-focused while making the value in the knowledge graph come from the distributed creation of knowledge. Why does this method work? The ontology in the knowledge graph is simple. Its value came from the users creating knowledge and adding it to the knowledge graph by their actions.
Additionally, when more users work on the platform, it follows more of them will take advantage of the knowledge graph recommendations. Consequently, the knowledge graph became richer as the analytical model is used more frequently. It is obvious that this method bypasses the ontology problem by placing the onus on user activities and the success of the users in creating useful knowledge.
The second innovative method for solving the ontology problem is to architect it using a distributed collection of smaller knowledge graphs. Similar to a standards-based approach, creating and integrating distributed microservices, with each contributing a small ontology along with data or computation, can facilitate a knowledge graph.
Furthermore, this approach requires a platform that understands microservice definitions such as the GraphQL definition released by Facebook and supported by the GraphQL Foundation. Here, the success and value of the knowledge graph achieved by the agglomeration of microservices are dependent on the same commodity the platform enables: the creation of valuable microservices by the customer.
Click here to read the article.
Please give your feedback on this article or share a similar story for publishing by clicking here.
