Domain-Adaptive Entity Resolution Algorithm Based on Semi-Supervised Learning

Strategic Insights -

Researchers at the National University of Defense Technology have developed a novel domain-adaptive entity resolution model that tackles the challenge of obtaining annotated data for entity resolution tasks. The proposed model utilizes semi-supervised learning to reduce the distributional difference between source and target domains, achieving comparable performance to supervised baseline models with only 20% to 40% of the labels required. The model's effectiveness was demonstrated through ablation experiments and comparison with unsupervised baseline models on 13 datasets from various domains.

Key Takeaways:

  • The proposed domain-adaptive entity resolution model achieves an average F1 score improvement of 2.84%, 9.16%, and 7.1% across multiple datasets compared to unsupervised baseline models.
  • The model achieves comparable performance to supervised baseline models with only 20% to 40% of the labels required.
  • Ablation experiments demonstrate the effectiveness of the proposed model, showing better entity resolution results in general.
  • The research proposes a novel semi-supervised learning approach to entity resolution, reducing the need for large amounts of annotated data.
  • The model's results are presented on 13 datasets from various domains, including algorithms, machine learning, and computer software development.

Statistics:

  • Average F1 score improvement: 2.84%, 9.16%, and 7.1% across multiple datasets.
  • Number of datasets used for testing: 13.
  • Number of labels required for comparable performance to supervised baseline models: 20% to 40%.
  • Relevant code available: [The relevant code is available[superscript]1)].

Sources:

  • Ding, H., Dai, C., (2024). Domain-adaptive Entity Resolution Algorithm Based on Semi-supervised Learning. Jisuanji kexue, 2024, 51(9):214-222.

https://doi-org.sdpl.idm.oclc.org/10.11896/jsjkx.230800102

  • National Key Laboratory of Information Systems Engineering, National University of Defense Technology, Changsha 410073, People's Republic of China.