Presentation 2021-12-13
Embedding Models for Logical Inference on Individuals and Subsumption Relations
Yukihiro Shiraishi, Ken Kaneiwa,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Knowledge graph embeddings are limited to predicting missing links. The embedding model EmEL++ can infer logical relationships in the Description Logic EL++. However, it does not achieve sufficient performance in inferences on individuals and binary relations. In this paper, we propose a revised model of EmEL++ by distinguishing individuals and classes in the loss functions for membership and subsumption relations. To enhance the embeddings of subsumption relations, we divide learning embeddings into the two steps of subsumption relations of concepts and roles and other relations. In the experiments, we show that the proposed embedding model outperforms EmEL++ and E2R in the inferences on individuals and binary relations using the LUBM benchmark.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) knowledge graph embedding / description logic / machine learning / inference
Paper # AI2021-4
Date of Issue 2021-12-06 (AI)

Conference Information
Committee AI
Conference Date 2021/12/13(1days)
Place (in Japanese) (See Japanese page)
Place (in English) Aimattain Hakata station
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair Yuichi Sei(Univ. of Electro-Comm.)
Vice Chair Yuko Sakurai(AIST) / Tadachika Ozono(Nagoya Inst. of Tech.)
Secretary Yuko Sakurai(Tokyo Univ. of Agriculture and Technology) / Tadachika Ozono(Toho Univ.)
Assistant Kazutaka Matsuzaki(Chuo Univ.)

Paper Information
Registration To Technical Committee on Artificial Intelligence and Knowledge-Based Processing
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Embedding Models for Logical Inference on Individuals and Subsumption Relations
Sub Title (in English)
Keyword(1) knowledge graph embedding
Keyword(2) description logic
Keyword(3) machine learning
Keyword(4) inference
1st Author's Name Yukihiro Shiraishi
1st Author's Affiliation The University of Electro-Communications(UEC)
2nd Author's Name Ken Kaneiwa
2nd Author's Affiliation The University of Electro-Communications(UEC)
Date 2021-12-13
Paper # AI2021-4
Volume (vol) vol.121
Number (no) AI-298
Page pp.pp.18-23(AI),
#Pages 6
Date of Issue 2021-12-06 (AI)