1. The document proposes a method to predict trust relationships between web users based on evidence theory. It uses user ratings of web content as evidence to infer trust, with each rating treated as a piece of evidence.
2. The method computes personalized trust recommendations by analyzing the provenance of existing trust annotations in social networks. It then uses the computed trust values to personalize websites, like recommending movies based on trust.
3. The approach integrates trust, provenance, and annotations on the semantic web to process information. Two applications are presented that illustrate using trust annotations and provenance for personalization.