Enhancing Collaborative Variational Autoencoder with Tag and Citation Information for Scientific Article Recommendation

• Ho Xanh

  University of Science VNUHCM

• 相澤 彰子

  国立情報学研究所

抄録

Hybrid methods such as collaborative deep learning (CDL) and collaborative variational autoencoder (CVAE) have become state-of-the-art methods in recommender systems for scienti_c articles. However, they typically use only information from titles and abstracts of arti-cles, and ignore potentially useful information in the tags and citations. Therefore, they may miss articles that contain vastly di_erent content from other articles, although those articles present the same topic. We addressed this problem by developing the CiT-CVAE model that consid- ers tag and citation information when providing recommendations. Our experimental results indicate that the proposed model achieves consis- tent improvement compared with CDL and CVAE.

収録刊行物

• Proceedings of Toward Effective Support for Academic Information Search Workshop

  Proceedings of Toward Effective Support for Academic Information Search Workshop 1 15-26, 2018-11-22

  九州大学大学院システム情報科学研究院

キーワード

• Recommender systems
• Scientific article recommendation
• Hybrid methods
• Citation information
• Tag information

詳細情報

• CRID: 1390853649416699520
• NII論文ID: 120006705765
• DOI: 10.5109/2230667
• HANDLE: 2324/2230667
• ISSN: 2435385X
• 本文言語コード: en
• データソース種別:
  • JaLC
  • IRDB
  • Crossref
  • CiNii Articles
  • KAKEN
• 抄録ライセンスフラグ: 使用可