Big Data challenges the customized information service mode in traditional enterprise decision support systems,
while application of keyword based search method brings down user decision making efficiency. A method has been
proposed to solve the problem by making the requirement generation and search process automatic, based on
previously constructed domain knowledge about the latent relations from decision task types to commonly required
information. It is verified through experiments that: 1) Efficiency of search process is improved as user time cost on
INF-REQ modeling is saved; 2) Precision of search results is improved, as the generated INF-REQ model is usually