Example of what I'm trying to achieve is in the image below. One "NODE" can have relation to multiple combinations of other nodes Ax and Bx. If I use first approach, I loose combination info (as in first case - is it A1+B1 or A1+B2?). In second approach - these red dots are aggregating nodes. Is there some construct like this in neo4j, or is it just standard node? Is there some other way that I could model this - without using these intermediary nodes?
Welcome to the community, Dragutin! Your question is very abstract, so maybe there is a better way to model the data in your particular use case, but in general the construct you describe is a hyperedge (i.e. an edge that connects more than 2 nodes) and the only way to model that in Neo4j is via intermediate (standard) nodes.
Thanks for confirmation. Well, I'm just categorizing some words - and those words have combination of different categories at the same time. It is more-less the same problem as I described it here just instead of a node you have "BOOK" and instead A1 you get "NOUN" and instead B1 you get "PLURAL" and instead B2 "SINGULAR" etc...
Do you think there could be some other way to do categorization like this?