Skip to content

Latest commit

 

History

History
93 lines (73 loc) · 3.11 KB

README.md

File metadata and controls

93 lines (73 loc) · 3.11 KB
Raw

[AAAI 24 (oral)] TEILP: Time Prediction over Knowledge Graphs via Logical Reasoning

This repository contains the code for the paper TEILP: Time Prediction over Knowledge Graphs via Logical Reasoning.

Introduction

This is a follow-up work of TILP: Differentiable Learning of Temporal Logical Rules on Knowledge Graphs. We convert TKGs into a temporal event knowledge graph (TEKG) which equips us to develop a differentiable random walk approach. We also introduce conditional probability density functions, associated with the logical rules involving the query interval, using which we arrive at the time prediction.

Quick Start

The structure of the file folder should be like

TEILP/
│
├── src/
│
├── data/
│
├── exps/
│
└── output/

Get into the path

cd src

Random walk:

# For wiki, YAGO:
python random_walk_for_interval_datasets.py --dataset {$dataset name}

# For icews14, icews05-15, gdelt100:
python random_walk_for_timestamp_datasets.py --dataset {$dataset name}

# Examples:
# python random_walk_for_interval_datasets.py --dataset wiki
# python random_walk_for_timestamp_datasets.py --dataset icews14

Rule learning:

python main.py --dataset {$dataset name} --train

# Examples:
# python main.py --dataset YAGO --train
# python main.py --dataset icews14 --train

Rule application:

python main.py --dataset {$dataset name} --test --from_model_ckpt {$path to saved model}

# Examples:
# python main.py --dataset YAGO --test --from_model_ckpt ../exps/YAGO_24-02-17-20-57/ckpt/model-30
# python main.py --dataset icews14 --test --from_model_ckpt ../exps/icews14_24-02-18-11-03/ckpt/model-30

Print rules:

# Find the rules and scores in output/{$dataset name}/rule_scores.
# Parameters:
# - rule_patterns: relations and temporal relations (Example in YAGO: query rel: 4, rule: 7 10 7 10 14 ukn af bf af bf; Translation: isMarriedTo(x, y, I_q) <- diedIn(x, e1, I1) and wasBornIn^{-1}(e1, e2, I2) and diedIn(e2, e3, I3) and wasBornIn^{-1}(e3, e4, I4) and isMarriedTo^{-1}(e4, y, I5) and unknown(I_q, I_1) and after(I_1, I_2) and before(I_2, I_3) and after(I_3, I_4) and before(I_4, I_5) and unknown(I_5, I_q))
# - rule_scores: score of the rules
# - refType_scores: score of reference types shared by all the rules (use which event in the body and its start time or end time)

Contact

If you have any inquiries, please feel free to raise an issue or reach out to [email protected].

Citation

@inproceedings{xiong2024teilp,
  title={Teilp: Time prediction over knowledge graphs via logical reasoning},
  author={Xiong, Siheng and Yang, Yuan and Payani, Ali and Kerce, James C and Fekri, Faramarz},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  number={14},
  pages={16112--16119},
  year={2024}
}