Building a simple recommendation engine in Neo4j

Graph databases like Neo4j are an excellent tool for creating recommendation engines. They allow us to examine a large context of a data point potentially comprising various data sources. Their powerful storage model is very well suited for applications where we want to analyze the direct surrounding of a node. If you would like understand what makes graphs so powerful in comparison with relational models, read here. In this article, I describe how we implemented a simple recommendation engine directly in Neo4j using only Cypher. The approach bases on basic NLP and simple conditional probabilities to find the most likely matching item.