采用元路径时效衰减和引用模式划分的学术引文推荐

发布时间：2018-06-12 17:35

本文选题：学术引文推荐 + 元路径　；参考：《西安交通大学学报》2017年07期

【摘要】：针对现有学术引文推荐算法中元路径特征无法衡量学术文献的时效性,且对元路径特征的利用和划分粒度较粗从而导致推荐精度不高的问题,提出了一种采用元路径时效衰减和引用模式划分的学术引文推荐方法。首先,利用元路径抽取丰富的引文特征,并在计算元路径特征时加入了时效衰减因子,提升了新发表文献的推荐精度;其次,提出了融合元路径特征的主题模型MpTM,该模型利用主题特征为文献划分引用模式,并联合学习文献的主题分布、引用模式和元路径特征权重,细化了元路径特征的粒度;最后,通过所有引用模式下的元路径特征加权值,为目标文献推荐学术引文。在AAN数据集上的实验结果表明:所提方法在准确率和召回率上平均提升约41.99%和22.43%,能够提升新发表文献和非权威文献的推荐精度,并能有效缓解引文链接的稀疏性问题。
[Abstract]:In view of the fact that the meta-path feature in the current academic citation recommendation algorithm can not measure the timeliness of academic literature, and the coarse granularity of the utilization and partition of meta-path features leads to the low recommendation accuracy. An academic citation recommendation method based on meta-path aging attenuation and citation model is proposed. Firstly, we use the meta-path to extract rich citation features, and add a time-dependent attenuation factor in the calculation of meta-path features, which improves the recommendation accuracy of the newly published literature. A topic model, MpTMM, which combines meta-path features, is proposed. The model uses topic features to divide reference patterns into reference patterns, and jointly studies the topic distribution, the weights of reference patterns and meta-path features to refine the granularity of meta-path features. This paper recommends academic citations for the target literature through the weighted values of meta path characteristics under all the citation modes. The experimental results on the AAN dataset show that the proposed method can improve the accuracy and recall rate by about 41.99% and 22.43% on average, which can improve the recommendation accuracy of the newly published and non-authoritative literatures, and can effectively alleviate the sparsity problem of citation links.
【作者单位】：西安交通大学软件学院;
【基金】：国家自然科学基金资助项目(61373046) 陕西省自然科学基金资助项目(S2015YFJM2129)
【分类号】：TP391.3

【参考文献】