User modeling

Preference model

Preference modeling refers to modeling users’ individual tastes and allows for a personalized experience. We model preferences as a Personal Knowledge Graph (PKG), where nodes can be either items or attributes. The preference model assumes a temporal ordering of the preference data and is built such that it remains consistent across simulations and assumes a temporal ordering.

Context model

In addition to preference and interaction modeling, we also model other user contexts, such as temporal and relational context. For example, during weekdays, a person might be busy with work and have other obligations that can change e.g., the information that they are willing to provide. On the other hand, during weekends, people are usually more relaxed and in a better mood in general, which could lead to the opposite effect.

Persona

Persona contains user variables, e.g., age, education level, etc. These characteristics also affect the behaviour and language usage of a person. People with higher education might use more formal or refined language. Age might also affect word choices and sentence structure.