Bayesian Inference with Complex Knowledge Graph Evidence

Authors: Armin Toroghi, Scott Sanner

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experimentally evaluate BIKG in incremental KGQA and interactive recommendation tasks demonstrating that it outperforms non-incremental methodologies and leads to better incorporation of conjunctive evidence vs. existing complex KGQA methods like CQD that leverage fuzzy T-norm operators.
Researcher Affiliation Academia Armin Toroghi1, Scott Sanner1,2 1Department of Mechanical and Industrial Engineering, University of Toronto 2Vector Institute of Artificial Intelligence, Toronto armin.toroghi@mail.utoronto.ca, ssanner@mie.utoronto.ca
Pseudocode Yes Algorithm 1: BIKG Algorithm
Open Source Code Yes 1https://github.com/atoroghi/BIKG
Open Datasets Yes We evaluate BIKG on three KGs: FB15k (Bordes et al. 2013), FB15k-237 (Toutanova and Chen 2015), and NELL995 (Xiong, Hoang, and Wang 2017). ... Movielens 20M (Harper and Konstan 2015) and LFM-1b (Schedl 2016).
Dataset Splits No The paper mentions 'training set of KG' and 'validation and test sets' when describing query extraction, but does not provide specific percentages or sample counts for train/validation/test splits of the KG data used for model training or evaluation.
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory) used to run the experiments.
Software Dependencies No The paper mentions the use of 'Simpl E KGE method' but does not provide specific version numbers for any software dependencies like programming languages, libraries, or frameworks used for implementation.
Experiment Setup No The paper provides the training objective for SimplE KGE (Equation 1) which includes a regularization hyperparameter λ, but it does not specify concrete values for this hyperparameter or other experimental setup details such as learning rate, batch size, or optimizer settings.