1 of 13

Exploiting Semi-Structured Information in Wikipedia for Knowledge Graph Construction

1

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

- joint work with Heiko Paulheim

2 of 13

MOTIVATION�LISTINGS & LIST PAGES

2

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

Lakes of Temagami

List of lakes of Powell County, Montana

Vinyl Williams

List of The Young and Restless characters

3 of 13

MOTIVATION�CATEGORIES, LIST CATEGORIES & LIST PAGES

3

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

4 of 13

OUTCOME (SO FAR)� v3.1

  • a rich ontology with�~1.3M classes & 3.2K properties

​

  • semantic class restrictions for�~500K classes�
  • new entities and facts from listings�~13.7M entities & 34.3M assertions

​

​

more information on http://caligraph.org�

​

​

4

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

clgo:Alternative_rock_song

clgo:Song

clgo:Nine_Inch_Nails_song

owl:Restriction

property = clgo:genre

value = clgr:Alternative_rock

owl:Restriction

property = clgo:artist

value = clgr:Nine_Inch_Nails

dbo:Song

clgo:Musical_work

5 of 13

OVERVIEW

  • (1+2) Learning a taxonomy from Wikipedia categories and list pages
  • (3) Learning axioms for categories

​

​

​

  • (4+5) Identifying and linking entities in Wikipedia listings
  • (6+7) Extracting novel information about listings in Wikipedia pages

​

​

​

​

5

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

6 of 13

LEARNING A TAXONOMY�FROM CATEGORIES, LIST CATEGORIES & LIST PAGES

6

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

7 of 13

LEARNING A TAXONOMY�FROM CATEGORIES, LIST CATEGORIES & LIST PAGES

7

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

  1. Identify lexical head
  2. Remove topic categories
  3. Remove invalid edges

​

​

​

  • Find synonyms and hypernyms between categories and list pages

​

​

​

​

  • Create classes and map to DBpedia

 

 

Nicolas Heist and Heiko Paulheim: Entity Extraction from Wikipedia List Pages, ESWC 2020

8 of 13

LEARNING AXIOMS FOR CATEGORIES�WITH Cat2Ax

8

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

Nicolas Heist and Heiko Paulheim: Uncovering the Semantics of Wikipedia Categories, ISWC 2019

 

 

 

9 of 13

LINKING ENTITIES IN WIKIPEDIA LISTINGS

9

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

9

Lakes of Temagami

List of lakes of Powell County, Montana

Vinyl Williams

List of The Young and Restless characters

10 of 13

LINKING ENTITIES IN WIKIPEDIA LISTINGS �WITH NASTyLinker (NIL-Aware Scalable Transformer-based Entity Linker)�

10

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

Nicolas Heist and Heiko Paulheim: NASTyLinker: NIL-Aware Scalable Transformer-based Entity Linker, ESWC 2023

11 of 13

INFORMATION EXTRACTION FOR LISTINGS ��

11

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

Nicolas Heist and Heiko Paulheim: Information Extraction from Co-Occurring Similar Entities, WWW 2021

…

LISTING 1

LISTING 2

LISTING 3

LISTING 4

LISTING 5

(artist, Axl_Rose)

(type, MusicalWork)

(type, Work)

(artist,�Guns_N’_Roses)

…

(starring, Axl_Rose)

 

 

…

12 of 13

KNOWLEDGE GRAPH EVALUATION �WITH KGrEaT (Knowledge Graph Evaluation via Downstream Tasks)�

12

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist

Nicolas Heist, Sven Hertling and Heiko Paulheim: KGrEaT: A Framework to Evaluate Knowledge Graphs via Downstream Tasks, CIKM 2023

TEASER

13 of 13

13

DBpedia Day @ SEMANTICS‘23 | Nicolas Heist