Ecole d'ingénieur et centre de recherche en Sciences du numérique

Inductive entity typing alignment

Rizzo, Giuseppe; van Erp, Marieke;Troncy, Raphaël

ISWC 2014, 1st International Workshop on Linked Data for Information Extraction (LD4IE 2014), October 20, 2014, Riva del Garda, Italy

Aligning named entity taxonomies for comparing or combining di erent named entity extraction systems is a dicult task. Often taxonomies are mapped manually onto each other or onto a standardized ontology but at the loss of subtleties between di erent class extensions and domain speci c uses of the taxonomy. In this paper, we present an approach and experiments for learning customized taxonomy alignments between di erent entity extractors for di erent domains. Our inductive data-driven approach recasts the alignment problem as a classi cation problem. We present experiments on two named entity recognition benchmark datasets, namely the CoNLL2003 newswire dataset and the MSM2013 microposts dataset. Our results show that the automatically induced mappings outperform manual alignments and are agnostic to changes in the extractor taxonomies, implying that alignments are highly contextual.

Document Bibtex

Titre:Inductive entity typing alignment
Type:Conférence
Langue:English
Ville:Riva del Garda
Pays:ITALIE
Date:
Département:Data Science
Eurecom ref:4400
Copyright: Copyright ESA. Personal use of this material is permitted. The definitive version of this paper was published in ISWC 2014, 1st International Workshop on Linked Data for Information Extraction (LD4IE 2014), October 20, 2014, Riva del Garda, Italy and is available at :
Bibtex: @inproceedings{EURECOM+4400, year = {2014}, title = {{I}nductive entity typing alignment}, author = {{R}izzo, {G}iuseppe and van {E}rp, {M}arieke and {T}roncy, {R}apha{\"e}l }, booktitle = {{ISWC} 2014, 1st {I}nternational {W}orkshop on {L}inked {D}ata for {I}nformation {E}xtraction ({LD}4{IE} 2014), {O}ctober 20, 2014, {R}iva del {G}arda, {I}taly}, address = {{R}iva del {G}arda, {ITALIE}}, month = {10}, url = {http://www.eurecom.fr/publication/4400} }
Voir aussi: