A generative adversarial network for single and multi-hop distributional knowledge base completion

Zia, Tehseen and Windridge, David ORCID logoORCID: https://orcid.org/0000-0001-5507-8516 (2021) A generative adversarial network for single and multi-hop distributional knowledge base completion. Neurocomputing, 461 . pp. 543-551. ISSN 0925-2312 [Article] (doi:10.1016/j.neucom.2021.04.128)

PDF - Final accepted version (with author's formatting)
Available under License Creative Commons Attribution-NonCommercial-NoDerivatives 4.0.

Download (1MB) | Preview
PDF (Preprint (pre-peer review)) - Draft pre-submission version (with author's formatting)
Download (1MB) | Preview


Knowledge bases (KBs) inherently lack reasoning ability, limiting their effectiveness for tasks such as question-answering and query expansion. Machine-learning is hence commonly employed for representation learning in order to learn semantic features useful for generalization. Most existing methods utilize discriminative models that require both positive and negative samples to learn a decision boundary. KBs, by contrast, contain only positive samples, necessitating that negative samples are generated by replacing the head/tail of predicates with randomly-chosen entities. They are thus frequently easily discriminable from positive samples, which can prevent learning of sufficiently robust classifiers. Generative models, however, do not require negative samples to learn the distribution of positive samples; stimulated by recent developments in Generative Adversarial Networks (GANs), we propose a novel framework, Knowledge Completion GANs (KCGANs), for competitively training generative link prediction models against discriminative belief prediction models. KCGAN thus invokes a game between generator-network G and discriminator-networkD in which G aims to understand underlying KB structure by learning to perform link prediction while D tries to gain knowledge about the KB by learning predicate/triplet classification. Two key challenges are addressed: 1) Classical GAN architectures’ inability to easily generate samples over discrete entities; 2) the inefficiency of softmax for learning distributions over large sets of entities. As a step toward full first-order logical reasoning we further extend KCGAN to learn multi-hop logical entailment relations between entities by enabling G to compose a multi-hop relational path between entities and D to discriminate between real and fake paths.

KCGAN is tested on benchmarks WordNet and FreeBase datasets and evaluated on link prediction and belief prediction tasks using MRR and HIT@10, achieving best-in-class performance.

Item Type: Article
Research Areas: A. > School of Science and Technology > Computer Science
Item ID: 33307
Notes on copyright: © 2021. The accepted manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/
Useful Links:
Depositing User: David Windridge
Date Deposited: 26 May 2021 07:41
Last Modified: 29 Nov 2022 17:42
URI: https://eprints.mdx.ac.uk/id/eprint/33307

Actions (login required)

View Item View Item


Activity Overview
6 month trend
6 month trend

Additional statistics are available via IRStats2.