With the development of information technology, the issue of trust becomes more and more important. In e-commerce or service-oriented environments, when there are a few sellers or service providers providing the same product/service, the buyer or service client would like to request the trust management authority to provide trust values of sellers or service providers, which are based on the ratings reflecting the quality of previous transactions. In addition, trust is context dependent, i.e. for different context of transactions, there are different factors influencing the trust result. In this paper, we propose a fuzzy comprehensive evaluation based method for building up a projection from the trust ratings in the transaction history of a service provider to an upcoming transaction depending on the similarity between previous transactions and the upcoming one, and the familiarity between each rater and the service client of the upcoming transaction. This process is termed as context based trust normalization. After trust normalization, normalized trust ratings are used for trust evaluation, the results of which would be closely bound to the upcoming transaction. Finally, we introduce the results of our conducted experiments to illustrate how our proposed method can detect some typical risks.