Presentation 2020-03-11
Knowledge Graph Completion by Separating Transition and Score Functions
Kenta Hama, Takashi Matsubara, Kuniaki Uehara,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) A knowledge graph is represented by a set of two entities and the relations, and used for various tasks such as information extraction, question answering, and sentence understanding. Since many knowledge graphs include missing relations and entities, Knowledge Graph Completion (KGC) is important to use for other tasks. Translation-based Models can predict missing entities by learning transition functions and embeddings of entities. However, many of these models are known that the accuracy of prediction is reduced in tasks such as Path Query Answering (PQA), which makes multiple transitions. On the other hand, if the transition embedding point is matched with the correct embedding point in order to prevent the loss of prediction accuracy, the embedding point having a one-to-many relationship will be collapsed into one point. In this study, we tried to solve the above problem by defining a transition function and an evaluation function of transition points separately in translation-based models. The proposed method improved accuracy in PQA and KGC.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Knowledge Graph / Link Prediction / Knowledge Graph Embedding
Paper # IBISML2019-41
Date of Issue 2020-03-03 (IBISML)

Conference Information
Committee IBISML
Conference Date 2020/3/10(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Kyoto University
Topics (in Japanese) (See Japanese page)
Topics (in English) Machine learning, etc.
Chair Hisashi Kashima(Kyoto Univ.)
Vice Chair Masashi Sugiyama(Univ. of Tokyo) / Koji Tsuda(Univ. of Tokyo)
Secretary Masashi Sugiyama(Nagoya Inst. of Tech.) / Koji Tsuda(AIST)
Assistant Tomoharu Iwata(NTT) / Shigeyuki Oba(Kyoto Univ.)

Paper Information
Registration To Technical Committee on Infomation-Based Induction Sciences and Machine Learning
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Knowledge Graph Completion by Separating Transition and Score Functions
Sub Title (in English)
Keyword(1) Knowledge Graph
Keyword(2) Link Prediction
Keyword(3) Knowledge Graph Embedding
1st Author's Name Kenta Hama
1st Author's Affiliation Kobe University(Kobe Univ.)
2nd Author's Name Takashi Matsubara
2nd Author's Affiliation Kobe University(Kobe Univ.)
3rd Author's Name Kuniaki Uehara
3rd Author's Affiliation Kobe University(Kobe Univ.)
Date 2020-03-11
Paper # IBISML2019-41
Volume (vol) vol.119
Number (no) IBISML-476
Page pp.pp.59-62(IBISML),
#Pages 4
Date of Issue 2020-03-03 (IBISML)