In view of the problem that the backtracking search method has a low efficiency of paper-generating under the condition of a large number of test questions, this paper proposes 2 improved algorithms: the target termination backtracking search method and the reduced depth backtracking search method. Through building the model of the improved backtracking search method paper-generating algorithm, the paper analyzed the applicability of the 2 improved algorithms in the paper-generating efficiency and verified the improved backtracking search method. The experimental results show that the 2 improved algorithms are obviously improved in the efficiency of the paper-generating, and each has its own characteristics and adaptability. If the amount of data is huge, the reduced depth backtracking search method is applicable, otherwise, the target termination backtracking search method is applied.