A reputation system computes and publishes reputation scores regarding any kind of entity (e.g. services or goods) within a community. Since the explosion of Internet the quantity of opinions regarding a particular item has increased dramatically, so the relevance of reputation systems is greater than ever. Nevertheless, experience shows that information held in a reputation centre is not fully reliable. The first challenge is to evaluate the trustfulness of the information. Secondly, a user who is looking for advice would prefer to give more importance to the opinions from people with a similar profile (age, gender, etc.) in order to get a customized advice. This work describes a framework that combines these two challenges in a single model. A user who is looking for advice will get a customized score for an entity based on how other users have reviewed this particular entity. In addition, the system considers the reputation of the reviewers and matches the similarity of each reviewer with the user looking for advice. The tool used to develop this model is subjective logic, a kind of probabilistic logic that allows expressing uncertainty in absence of explicit belief. Finally, we test the model with some invented scenarios to validate it.

Outgoing