Abstract: In order to discover the relationship between literature academic characters and the novelty of sci-tech project retrieval, this paper studied the dependence influence between the project novelty of scitech novelty retrieval and the four key academic characters, namely, object, content, method and result. The multiple-regression analysis method, 54 agronomy articles, and five agronomy projects randomly selected were used in this paper. The academic character similarity and correlation of sci-tech novelty retrieval project in those articles was respectively evaluated and scored by 10 professional novelty researchers. According to the results of the grading score, articles closely related were extracted. A total of 900 sets of data were obtained, and three different types of correlations were found, leading to a multiple regression combination model that was tested and verified afterwards.

Key words: Literature academic characters, Sci-tech novelty retrieval, Novelty, MLR(multiple linear regression), Interval estimation