Design a system that recommends the next Confluence pages a user should view.

A graph-based approach is well-suited for recommending Confluence pages due to the inherent interconnectedness of knowledge. By modeling pages as nodes and content-based references as edges in a graph database, we can effectively capture relationships. This allows for recommendations based on traversal paths, such as pages frequently linked from a user's current or recently viewed content. This method leverages the explicit structure of Confluence, offering more contextually relevant suggestions than a purely collaborative filtering or two-tower model might provide on its own.