IRI stands for Internationalized Resource Identifier. In the RDF space IRIs are used as “names”, or an equivalent of “IDs”, for graph nodes.
IRIs used in popular ontologies look a lot like most URLs we are used to. This is because IRIs are defined as a superset of URLs (every URL is an IRI, but not the other way around) and have the same structure with schemes (e.g., “http”), paths and fragments. The main difference is that IRIs can contain characters not frequently used in English, but which appear in other languages (hence “internationalized”), such as é, ü or ç.
Suppose you own the website “www.example.com”. You can build a knowledge graph with IRIs such as “http://www.example.com/person_alice” and have information about each subject (here: Alice) accessible online at the same address. This is the idea at the foundation of the “Semantic Web”, a concept which focuses on creating a direct link between the data stored in databases and available on the internet. That being said, following these practices is not necessary for building effective, intelligent applications with graph databases.