Combining embedding methods for a word intrusion task

Finn Årup Nielsen; Lars Kai Hansen

Combining embedding methods for a word intrusion task

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

41 Downloads (Pure)

Abstract

We report a new baseline for a Danish word intrusion task by combining pre-trained offthe-shelf word, subword and knowledge graph embedding models. We test fastText, Byte-Pair Encoding, BERT and the knowledge graph embedding in Wembedder, finding fastText as the individual model with the superior performance, while a simple combination of the fastText with other models can slightly improve the accuracy of finding the odd-one-out words in the word
intrusion task.

Original language	English
Title of host publication	Proceedings of the 15^th Conference on Natural Language Processing
Publisher	Association for Computational Linguistics
Publication date	2019
Pages	237-240
Publication status	Published - 2019
Event	15^th Conference on Natural Language Processing - Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany Duration: 9 Oct 2019 → 11 Oct 2019 https://2019.konvens.org/

Conference

Conference	15^th Conference on Natural Language Processing
Location	Friedrich-Alexander-Universität Erlangen-Nürnberg
Country/Territory	Germany
City	Erlangen
Period	09/10/2019 → 11/10/2019
Internet address	https://2019.konvens.org/

Access to Document

FulltextFinal published version, 238 KB

OpenUrl availability

Full text

Cite this

@inproceedings{be71173384a04097b7190a50c7f156f7,

title = "Combining embedding methods for a word intrusion task",

abstract = "We report a new baseline for a Danish word intrusion task by combining pre-trained offthe-shelf word, subword and knowledge graph embedding models. We test fastText, Byte-Pair Encoding, BERT and the knowledge graph embedding in Wembedder, finding fastText as the individual model with the superior performance, while a simple combination of the fastText with other models can slightly improve the accuracy of finding the odd-one-out words in the wordintrusion task. ",

author = "Nielsen, {Finn {\AA}rup} and Hansen, {Lars Kai}",

year = "2019",

language = "English",

pages = "237--240",

booktitle = "Proceedings of the 15th Conference on Natural Language Processing",

publisher = "Association for Computational Linguistics",

note = "15<sup>th</sup> Conference on Natural Language Processing, KONVENS 2019 ; Conference date: 09-10-2019 Through 11-10-2019",

url = "https://2019.konvens.org/",

}

TY - GEN

T1 - Combining embedding methods for a word intrusion task

AU - Nielsen, Finn Årup

AU - Hansen, Lars Kai

PY - 2019

Y1 - 2019

N2 - We report a new baseline for a Danish word intrusion task by combining pre-trained offthe-shelf word, subword and knowledge graph embedding models. We test fastText, Byte-Pair Encoding, BERT and the knowledge graph embedding in Wembedder, finding fastText as the individual model with the superior performance, while a simple combination of the fastText with other models can slightly improve the accuracy of finding the odd-one-out words in the wordintrusion task.

AB - We report a new baseline for a Danish word intrusion task by combining pre-trained offthe-shelf word, subword and knowledge graph embedding models. We test fastText, Byte-Pair Encoding, BERT and the knowledge graph embedding in Wembedder, finding fastText as the individual model with the superior performance, while a simple combination of the fastText with other models can slightly improve the accuracy of finding the odd-one-out words in the wordintrusion task.

M3 - Article in proceedings

SP - 237

EP - 240

BT - Proceedings of the 15th Conference on Natural Language Processing

PB - Association for Computational Linguistics

T2 - 15<sup>th</sup> Conference on Natural Language Processing

Y2 - 9 October 2019 through 11 October 2019

ER -

Combining embedding methods for a word intrusion task

Abstract

Conference

Access to Document

OpenUrl availability

Fingerprint

Cite this