Improving computational trust representation based on Internet auction traces

► We study a large trace of textual feedbacks from a reputation system of an Internet auction. ► We discover that users systematically try to add information in the textual comments. ► Text-mining and NLP approaches reveal a taxonomy of non-positive feedbacks. ► An importance order exists on the categories of non-positive behavior. ► A new computational trust representation that extends eBay's DSR is proposed and evaluated.