[目的/意義]科技文獻推薦是指根據學者的研究興趣,自動地為其推薦文獻資源。借助于文獻推薦,學者可以快速發現優質文獻,提高論文的撰寫效率。[方法/過程]首先,獲取學者已經發表的學術論文,以此為依據分析其研究興趣;其次,分別從學者研究興趣與待推薦文獻的語義相關性、待推薦文獻集合的主題多樣性以及影響力3個維度為待推薦文獻建模;最后,綜合考慮這3個因素為學者推薦科技文獻。[結果/結論]實驗結果表明,與傳統的推薦模型相比,文章提出的融合主題多樣性與影響力的科技文獻推薦算法能夠更好地刻畫待推薦文獻的特征,進一步提高文獻推薦算法的科學性和實用性。

[Purpose/significance] Scientific and technical( S&T) literature recommendation usually refers to automatically recommending the most valuable literature to scholars based on their research interests. With the help of paper recommendation system,scholars can get valuable papers quickly,which can help them to improve writing efficiency. [Method/process] In this research,scholars' research interests are extracted from their published papers. Then,the relevance between the scholars' research interests,the diversity and authority of candidate papers are deeply observed. Finally,these three important factors are combined in an integrated model to recommend papers to the scholars. [Result/conclusion] Experiments show that compared with the Vector Space Model and Latent Dirichlet Allocation,the proposed S&T literature recommendation method which integrates topic diversity and influence can describe the characteristics of candidate papers better.