Knowledge Graph Completion DataSet

    Refined Training and Test dataset for Knowledge Graph Model

Data overview

    This repository contains refinement dataset to train and evaluate knowledge graph model.
    If you have any questions or comments, please fell free to contact us by email [leewangon@gmail.com].

Data format

   This data is represented by a sequence of the following n-triple format.
   n-triple - a format for storing and transmitting data. 
            - a line-based, plain text serialisation format for RDF (Resource Description Framework) graphs.
            - a subset of the Turtle (Terse RDF Triple Language) format.
   <subject, relation, object>

Data Set

	Kor-KB	NELL-995	FB15K-237	NDSL	DBpedia	DBpia	KData
Size	42MB	26MB	41MB	26MB	711MB	64MB	139MB
Triples	1,315,146	308,426	544,230	221,253	14,000,000	912,412	2,776,394
Entities	488,926	63,917	14,505	246,850	4,250,000	409,693	1,140,000
Relations	157	396	237	5	717	4	10,409
Classes	921	267	354	5	451	6	113

Kor-KB

Task #1 : nationality

Task #2 : job

Task #3 : employer

NELL-995 Link

FB15K-237 Link

NDSL Link

DBpedia Link

DBpia Link

KData Link

The institute to construct dataset

• The AI Lab in Soongsil University

Citation

    @article{jagvaral2020path,
      title={Path-based reasoning approach for knowledge graph completion using CNN-BiLSTM with attention mechanism},
      author={Jagvaral, Batselem and Lee, Wan-Kon and Roh, Jae-Seung and Kim, Min-Sung and Park, Young-Tack},
      journal={Expert Systems with Applications},
      volume={142},
      pages={112960},
      year={2020},
      publisher={Elsevier}
    }