This paper focuses on the problems concerning soft information management in construction projects. Firstly, it is stated in the paper that soft information managements pertains to uncertain language multi-attribute group decision- making problem of incomplete information. Further, a decision model regarding soft information management in construction projects is built on the foundation of 1) the process of linguistic variables and 2) the uncertain language multi-attribute group decision-making model based on Dempster-Shafer theory. Soft information is difficult to quantify and can easily get lost when spreading among different units and the previous mentioned method can be better applied to solve these problems. In the end of this paper, empirical study is taken to confirm the efficiency and scientificalness of this method.