Reproduce “Question Answering Over Temporal Knowledge Graphs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Apoorv Saxena, Soumen Chakrabarti, Partha Talukdar
Researchers introduced CRONQUESTIONS, the largest-known dataset for question answering over temporal knowledge graphs, comprising 410,000 question-answer pairs that require temporal reasoning. They also developed CRONKGQA, a transformer-based model that achieved a 120% accuracy increase over baselines, demonstrating the effectiveness of integrating temporal knowledge graph embeddings for answering time-sensitive questions.
Key insights
- Large-scale temporal KGQA datasets are crucial for training robust models, as existing language models and non-temporal KGQA methods perform poorly on time-sensitive queries without explicit temporal guidance.
- Integrating explicit temporal knowledge graph embeddings into transformer-based language models significantly enhances their ability to perform temporal reasoning and retrieve time-dependent facts.
- While the proposed methods are highly effective for simple temporal reasoning, complex multi-step temporal inference (e.g., 'Before/After' relationships) remains a significant challenge requiring further research.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.