Elamur Research⌕
☰
home / papers / 2024 / TILP: Differentiable Learning of Temporal Logical Rules on Knowledge Graphs

TILP: Differentiable Learning of Temporal Logical Rules on Knowledge Graphs

Paper · 2024 · Siheng Xiong, Yuan Yang, Faramarz Fekri et al.

Summary

Temporal knowledge graphs (tKGs) attach validity intervals to facts, which makes link prediction harder than on static graphs. Embedding models are opaque and depend on absolute timestamps. Rule miners such as TLogic and StreamLearner score each rule separately by counting paths, so they miss interactions and similarities between rules and work with timestamps rather than intervals.

TILP is a differentiable framework that learns temporal logical rules. It uses constrained random walks that separate Markovian constraints (predicates, and relations to the query interval) from non-Markovian ones (pairwise relations between body intervals). Allen's 13 interval relations are reduced to three classes: before, touching and after. A separate module models temporal features such as recurrence, ordering, gaps between relation pairs and duration.

On WIKIDATA12k and YAGO11k, TILP matches or beats state-of-the-art baselines and explains each prediction with rules. Its advantage is larger when training data is scarce, imbalanced, or drawn from a different time period than the test data. This makes it attractive for interpretable reasoning over evolving facts.

Key points

  • Benchmarks: WIKIDATA12k and YAGO11k, both interval-based tKGs with yearly resolution.
    • They contain time-sensitive relations such as "member of sports team" and "playsFor".
  • Baselines: HyTE, TA-ComplEx, TNTComplEx, DE-SimplE, TimePlex and TLogic.
  • Harder settings:
    • Reduced training sets: repeated over 5 rounds.
    • Biased data: 50% of one relation's training edges removed.
    • Time shift: WIKIDATA12k splits by start year into train [0, 2008], validation [2008, 2012] and test [2012, 2018].
  • Why it generalizes: rules rely on relative temporal relations and relation-specific attention, not on absolute timestamps or shared entity embeddings.
  • Cost: with maximum rule length L=5 on a 4-CPU machine, rule search takes about 1741 s on WIKIDATA12k and 572 s on YAGO11k. Code is available at github.com/xiongsiheng/TILP.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Georgia Institute of TechnologyAtlanta · United States133 works · 24,497 citations

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 works

🧭 Topics 1Research subjects and languages the work is about.Knowledge graphs & ontologies · 192 works

Cite

Cite as Siheng Xiong, Yuan Yang, Faramarz Fekri and James Clayton Kerce. 2024. TILP: Differentiable Learning of Temporal Logical Rules on Knowledge Graphs. arXiv:2402.12309.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2402.12309 · arXiv 2402.12309

@misc{xiong2024tilp,
  title = {{TILP: Differentiable Learning of Temporal Logical Rules on Knowledge Graphs}},
  author = {Siheng Xiong, Yuan Yang, Faramarz Fekri and James Clayton Kerce},
  year = {2024},
  doi = {10.48550/arxiv.2402.12309},
  eprint = {2402.12309},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2402.12309}
}

References

29 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. James F Allen. Maintaining knowledge about temporal intervals. Communications of the ACM, 26 (11):832–843, 1983. find
  2. Ranran Bian, Yun Sing Koh, Gillian Dobbie, and Anna Divoli. Network embedding and change modeling in dynamic heterogeneous networks. In Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 861–864, 2019. find
  3. Shib Sankar Dasgupta, Swayambhu Nath Ray, and Partha Talukdar. HyTE: Hyperplane-based temporally aware knowledge graph embedding. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pp. 2001–2011, Brussels, Belgium, October-November 2018. Association for Comput… find
  4. Richard Evans and Edward Grefenstette. Learning explanatory rules from noisy data. Journal of Artificial Intelligence Research, 61:1–64, 2018. find
  5. Alberto Garcı́a-Durán, Sebastijan Dumančić, and Mathias Niepert. Learning sequence encoders for temporal knowledge graph completion. arXiv preprint arXiv:1809.03202, 2018. find
  6. Rishab Goel, Seyed Mehran Kazemi, Marcus Brubaker, and Pascal Poupart. Diachronic embedding for temporal knowledge graph completion, 2019. URL https://arxiv.org/abs/1907. 03143. ↗
  7. Vinh Thinh Ho, Daria Stepanova, Mohamed H Gad-Elrab, Evgeny Kharlamov, and Gerhard Weikum. Rule learning from knowledge graphs guided by embedding models. In International Semantic Web Conference, pp. 72–90. Springer, 2018. find
  8. Prachi Jain, Sushant Rathi, Soumen Chakrabarti, et al. Temporal knowledge base completion: New algorithms and evaluation protocols. arXiv preprint arXiv:2005.05035, 2020. find
  9. Tingsong Jiang, Tianyu Liu, Tao Ge, Lei Sha, Baobao Chang, Sujian Li, and Zhifang Sui. Towards time-aware knowledge graph completion. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pp. 1715–1724, 2016. find
  10. Timothée Lacroix, Guillaume Obozinski, and Nicolas Usunier. Tensor decompositions for temporal knowledge base completion, 2020. URL https://arxiv.org/abs/2004.04926. ↗
  11. Jennifer Lautenschlager, Steve Shellman, and Michael Ward. ICEWS Event Aggregations, 2015. URL https://doi.org/10.7910/DVN/28117. — [doi:10.7910/dvn/28117](https://doi.org/10.7910/dvn/28117) ↗
  12. Julien Leblay and Melisachew Wudage Chekol. Deriving validity time in knowledge graph. In Companion Proceedings of the The Web Conference 2018, pp. 1771–1776, 2018. find
  13. Kalev Leetaru and Philip A Schrodt. Gdelt: Global data on events, location, and tone, 1979–2012. In ISA annual convention, volume 2, pp. 1–49. find
  14. Yushan Liu, Yunpu Ma, Marcel Hildebrandt, Mitchell Joblin, and Volker Tresp. Tlogic: Temporal logical rules for explainable link forecasting on temporal knowledge graphs, 2021. URL https://arxiv.org/abs/2112.08025. — [arXiv:2112.08025](https://arxiv.org/abs/2112.08025) ↗
  15. Sedigheh Mahdavi, Shima Khoshraftar, and Aijun An. dynnode2vec: Scalable dynamic network embedding. In 2018 IEEE international conference on big data (Big Data), pp. 3762–3765. find
  16. Farzaneh Mahdisoltani, Joanna Biega, and Fabian Suchanek. Yago3: A knowledge base from multilingual wikipedias. In 7th biennial conference on innovative data systems research. CIDR Conference, 2014.
  17. Robin Manhaeve, Sebastijan Dumancic, Angelika Kimmig, Thomas Demeester, and Luc De Raedt. Deepproblog: Neural probabilistic logic programming. Advances in Neural Information Processing Systems, 31, 2018.
  18. Christian Meilicke, Melisachew Wudage Chekol, Manuel Fink, and Heiner Stuckenschmidt. Reinforced anytime bottom up rule learning for knowledge graph completion. arXiv preprint arXiv:2004.04412, 2020. find
  19. Mehrnoosh Mirtaheri, Mohammad Rostami, Xiang Ren, Fred Morstatter, and Aram Galstyan. One-shot learning for temporal knowledge graphs. arXiv preprint arXiv:2010.12144, 2020. — [arXiv:2010.12144](https://arxiv.org/abs/2010.12144) find
  20. Giang H Nguyen, John Boaz Lee, Ryan A Rossi, Nesreen K Ahmed, Eunyee Koh, and Sungchul Kim. Dynamic network embeddings: From random walks to temporal random walks. In 2018 IEEE International Conference on Big Data (Big Data), pp. 1085–1092. find
  21. Pouya Ghiasnezhad Omran, Kewen Wang, and Zhe Wang. Learning temporal rules from knowledge graph streams. In AAAI Spring Symposium: Combining Machine Learning with Knowledge Engineering, 2019. find
  22. Ali Payani and Faramarz Fekri. Inductive logic programming via differentiable deep neural logic networks, 2019. URL https://arxiv.org/abs/1906.03523. — [arXiv:1906.03523](https://arxiv.org/abs/1906.03523) ↗
  23. Tim Rocktäschel and Sebastian Riedel. End-to-end differentiable proving. Advances in neural information processing systems, 30, 2017. find
  24. Théo Trouillon, Johannes Welbl, Sebastian Riedel, Éric Gaussier, and Guillaume Bouchard. Complex embeddings for simple link prediction. In International conference on machine learning, pp. 2071–2080. PMLR, 2016.
  25. Po-Wei Wang, Daria Stepanova, Csaba Domokos, and J Zico Kolter. Differentiable learning of numerical rules in knowledge graphs. In International Conference on Learning Representations, 2019. find
  26. Leon Weber, Pasquale Minervini, Jannes Münchmeyer, Ulf Leser, and Tim Rocktäschel. Nlprolog: Reasoning with weak unification for question answering in natural language. arXiv preprint arXiv:1906.06187, 2019. find
  27. Yonghui Xu, Shengjie Sun, Huiguo Zhang, Chang’an Yi, Yuan Miao, Dong Yang, Xiaonan Meng, Yi Hu, Ke Wang, Huaqing Min, et al. Time-aware graph embedding: A temporal smoothness and task-oriented approach. ACM Transactions on Knowledge Discovery from Data (TKDD), 16(3): 1–23, 2021. find
  28. Fan Yang, Zhilin Yang, and William W. Cohen. Differentiable learning of logical rules for knowledge base reasoning, 2017. URL https://arxiv.org/abs/1702.08367. ↗
  29. Wen Zhang, Bibek Paudel, Liang Wang, Jiaoyan Chen, Hai Zhu, Wei Zhang, Abraham Bernstein, and Huajun Chen. Iteratively learning embeddings and rules for knowledge graph reasoning. In The World Wide Web Conference, pp. 2366–2377, 2019. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/xiong-2024-tilp-differentiable-learning-temporal-logical/