Interpreting Attention Models with Human Visual Attention in Machine Reading Comprehension

Vorschaubild
Datum
2020
Autor:innen
Sood, Ekta
Tannert, Simon
Bulling, Andreas
Vu, Ngoc Thang
Herausgeber:innen
Kontakt
ISSN der Zeitschrift
eISSN
item.preview.dc.identifier.isbn
Bibliografische Daten
Verlag
Schriftenreihe
URI (zitierfähiger Link)
DOI (zitierfähiger Link)
ArXiv-ID
Internationale Patentnummer
EU-Projektnummer
Projekt
Open Access-Veröffentlichung
Sammlungen
Gesperrt bis
Titel in einer weiteren Sprache
Forschungsvorhaben
Organisationseinheiten
Zeitschriftenheft
Publikationstyp
Beitrag zu einem Konferenzband
Publikationsstatus
Published
Erschienen in
Proceedings of the 24th Conference on Computational Natural Language Learning / Fernández, Raquel; Linzen, Tal (Hrsg.). - Stroudsburg, PA : ACL, 2020. - S. 12-25. - ISBN 978-1-952148-63-7
Zusammenfassung
While neural networks with attention mechanisms have achieved superior performance on many natural language processing tasks, it remains unclear to which extent learned attention resembles human visual attention. In this paper, we propose a new method that leverages eye-tracking data to investigate the relationship between human visual attention and neural attention in machine reading comprehension. To this end, we introduce a novel 23 participant eye tracking dataset - MQA-RC, in which participants read movie plots and answered pre-defined questions. We compare state of the art networks based on long short-term memory (LSTM), convolutional neural models (CNN) and XLNet Transformer architectures. We find that higher similarity to human attention and performance significantly correlates to the LSTM and CNN models. However, we show this relationship does not hold true for the XLNet models – despite the fact that the XLNet performs best on this challenging task. Our results suggest that different architectures seem to learn rather different neural attention strategies and similarity of neural to human attention does not guarantee best performance.
Zusammenfassung in einer weiteren Sprache
Fachgebiet (DDC)
400 Sprachwissenschaft, Linguistik
Schlagwörter
Konferenz
24th Conference on Computational Natural Language Learning (CoNLL) (online), 19. Nov. 2020 - 20. Nov. 2020
Rezension
undefined / . - undefined, undefined. - (undefined; undefined)
Zitieren
ISO 690SOOD, Ekta, Simon TANNERT, Diego FRASSINELLI, Andreas BULLING, Ngoc Thang VU, 2020. Interpreting Attention Models with Human Visual Attention in Machine Reading Comprehension. 24th Conference on Computational Natural Language Learning (CoNLL) (online), 19. Nov. 2020 - 20. Nov. 2020. In: FERNÁNDEZ, Raquel, ed., Tal LINZEN, ed.. Proceedings of the 24th Conference on Computational Natural Language Learning. Stroudsburg, PA:ACL, pp. 12-25. ISBN 978-1-952148-63-7. Available under: doi: 10.18653/v1/2020.conll-1.2
BibTex
@inproceedings{Sood2020Inter-59698,
  year={2020},
  doi={10.18653/v1/2020.conll-1.2},
  title={Interpreting Attention Models with Human Visual Attention in Machine Reading Comprehension},
  isbn={978-1-952148-63-7},
  publisher={ACL},
  address={Stroudsburg, PA},
  booktitle={Proceedings of the 24th Conference on Computational Natural Language Learning},
  pages={12--25},
  editor={Fernández, Raquel and Linzen, Tal},
  author={Sood, Ekta and Tannert, Simon and Frassinelli, Diego and Bulling, Andreas and Vu, Ngoc Thang}
}
RDF
<rdf:RDF
    xmlns:dcterms="http://purl.org/dc/terms/"
    xmlns:dc="http://purl.org/dc/elements/1.1/"
    xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
    xmlns:bibo="http://purl.org/ontology/bibo/"
    xmlns:dspace="http://digital-repositories.org/ontologies/dspace/0.1.0#"
    xmlns:foaf="http://xmlns.com/foaf/0.1/"
    xmlns:void="http://rdfs.org/ns/void#"
    xmlns:xsd="http://www.w3.org/2001/XMLSchema#" > 
  <rdf:Description rdf:about="https://kops.uni-konstanz.de/server/rdf/resource/123456789/59698">
    <dc:creator>Vu, Ngoc Thang</dc:creator>
    <dcterms:rights rdf:resource="https://rightsstatements.org/page/InC/1.0/"/>
    <dc:contributor>Tannert, Simon</dc:contributor>
    <dspace:isPartOfCollection rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/45"/>
    <dc:creator>Sood, Ekta</dc:creator>
    <dcterms:issued>2020</dcterms:issued>
    <dcterms:hasPart rdf:resource="https://kops.uni-konstanz.de/bitstream/123456789/59698/1/Sood_2-1idyw6jldeod6.pdf"/>
    <dc:creator>Tannert, Simon</dc:creator>
    <dc:date rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2023-01-12T14:40:30Z</dc:date>
    <dcterms:isPartOf rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/45"/>
    <dcterms:abstract xml:lang="eng">While neural networks with attention mechanisms have achieved superior performance on many natural language processing tasks, it remains unclear to which extent learned attention resembles human visual attention. In this paper, we propose a new method that leverages eye-tracking data to investigate the relationship between human visual attention and neural attention in machine reading comprehension. To this end, we introduce a novel 23 participant eye tracking dataset - MQA-RC, in which participants read movie plots and answered pre-defined questions. We compare state of the art networks based on long short-term memory (LSTM), convolutional neural models (CNN) and XLNet Transformer architectures. We find that higher similarity to human attention and performance significantly correlates to the LSTM and CNN models. However, we show this relationship does not hold true for the XLNet models – despite the fact that the XLNet performs best on this challenging task. Our results suggest that different architectures seem to learn rather different neural attention strategies and similarity of neural to human attention does not guarantee best performance.</dcterms:abstract>
    <dcterms:title>Interpreting Attention Models with Human Visual Attention in Machine Reading Comprehension</dcterms:title>
    <dspace:hasBitstream rdf:resource="https://kops.uni-konstanz.de/bitstream/123456789/59698/1/Sood_2-1idyw6jldeod6.pdf"/>
    <dc:contributor>Vu, Ngoc Thang</dc:contributor>
    <dcterms:available rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2023-01-12T14:40:30Z</dcterms:available>
    <dc:contributor>Sood, Ekta</dc:contributor>
    <dc:language>eng</dc:language>
    <dc:contributor>Bulling, Andreas</dc:contributor>
    <dc:contributor>Frassinelli, Diego</dc:contributor>
    <dc:creator>Bulling, Andreas</dc:creator>
    <foaf:homepage rdf:resource="http://localhost:8080/"/>
    <dc:creator>Frassinelli, Diego</dc:creator>
    <void:sparqlEndpoint rdf:resource="http://localhost/fuseki/dspace/sparql"/>
    <bibo:uri rdf:resource="https://kops.uni-konstanz.de/handle/123456789/59698"/>
    <dc:rights>terms-of-use</dc:rights>
  </rdf:Description>
</rdf:RDF>
Interner Vermerk
xmlui.Submission.submit.DescribeStep.inputForms.label.kops_note_fromSubmitter
Kontakt
URL der Originalveröffentl.
Prüfdatum der URL
Prüfungsdatum der Dissertation
Finanzierungsart
Kommentar zur Publikation
Allianzlizenz
Corresponding Authors der Uni Konstanz vorhanden
Internationale Co-Autor:innen
Universitätsbibliographie
Ja
Begutachtet