电子科技 ›› 2023, Vol. 36 ›› Issue (4): 59-64.doi: 10.16180/j.cnki.issn1007-7820.2023.04.008

• • 上一篇    下一篇

一种基于多任务学习的科学文献推荐算法

白莹琦1,帕丽旦·吐尔逊2   

  1. 1.西北大学 图书馆,陕西 西安,710127
    2.新疆师范大学,新疆 乌鲁木齐,830017
  • 收稿日期:2021-11-24 出版日期:2023-04-15 发布日期:2023-04-21
  • 作者简介:白莹琦(1983-)女,硕士研究生。研究方向:推荐系统。
  • 基金资助:
    国家社会科学基金项目(15CXW016);2019西北大学教育信息化研究项目(2019NWUXXH14);陕西省大学生创新创业训练计划项目(S202010697513);陕西省大学生创新创业训练计划项目(S202110697529)

A Scientific Literature Recommendation Method Based on Multi-Task Learning

BAI Yingqi1,PALIDAN·Tuerxun 2   

  1. 1. Library,Northwest University,Xi’an 710127,China
    2. Xinjiang Normal University,Urumqi 830017,China
  • Received:2021-11-24 Online:2023-04-15 Published:2023-04-21
  • Supported by:
    National Social Science Funds of China(15CXW016);Northwest University Education Informatization Research Project(2019NWUXXH14);Shaanxi Provincial Innovation and Entrepreneurship Training Program for College Students(S202010697513);Shaanxi Provincial Innovation and Entrepreneurship Training Program for College Students(S202110697529)

摘要:

传统推荐算法通过主题模型或者词语向量化的平均值对文本内容进行映射。针对现有方法不能充分利用文本信息或忽略词序信息这一问题,文中面向科学文献,提出了一种多任务学习推荐方法。该方法基于多任务学习框架,设计编码器并搭建了GL模型。该模型被训练为内容推荐与文本元数据预测的组合,可改善传统协同过滤的稀疏性问题,使得协同过滤模型正则化。最后,分别在公开数据集与私有数据集上进行了评估测试,结果表明所提方法性能优于现有的经典方法。

关键词: 推荐系统, 深度学习, 神经网络, 多任务学习, 协同过滤, 门控递归单元, 协同主题回归, 编码器

Abstract:

Traditional recommendation algorithms map text content through topic model or mean value of word vectorization. For the issue that existing methods cannot make full use of text information or ignore word order information, this study proposes a multi-task learning recommendation method for scientific literature. Based on the multi-task learning framework, an encoder is designed and a GL model is established. The GL model is trained to combine content recommendation and text metadata prediction, which improves the sparsity of traditional collaborative filtering and regularizes the collaborative filtering model. Finally, an evaluation test is carried out on public and private data sets respectively, and the superiority of the proposed method is demonstrated by comparing with the existing classical methods.

Key words: recommendation system, deep learning, neural networks, multi-task learning, collaborative filtering, gated recurrent unit, collaborative topic regression, encoder

中图分类号: 

  • TP391.3