[paper reading][EMNLP 2018] Learning Sequence Encoders for Temporal Knowledge Graph Completion

  • EMNLP 2018
  • https://arxiv.org/pdf/1809.03202.pdf
  • link prediction in knowledge graphs
  • static multi-relational data? hold for a time!
  • LSTM: a sequence for a predicate (with timestamp)

1 Introduction

  • triplet
  • completion: s, p, ?; ?, p, o
  • time info
    • s, p, o, time predicate, timestamp
    • sparse, irregular

3 Time-Aware Representations

  • learn a scoring function \(f\)
  • embedding of s,p,o
    • then \(f(s,p,o)\)
    • TransE, distMult
    • \(p_{seq}\)!
      [paper reading][EMNLP 2018] Learning Sequence Encoders for Temporal Knowledge Graph Completion
  • LSTMs
上一篇:论文阅读:ExCAR: Event Graph Knowledge Enhanced Explainable Causal Reasoning


下一篇:Proj AutoWriter Paper Reading: Knowledge-based type-aware neural text-to-sql generation