{"id":"https://openalex.org/W2949603537","doi":"https://doi.org/10.18653/v1/w19-4808","title":"Analyzing the Structure of Attention in a Transformer Language Model","display_name":"Analyzing the Structure of Attention in a Transformer Language Model","publication_year":2019,"publication_date":"2019-01-01","ids":{"openalex":"https://openalex.org/W2949603537","doi":"https://doi.org/10.18653/v1/w19-4808","mag":"2949603537"},"language":"en","primary_location":{"id":"doi:10.18653/v1/w19-4808","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/w19-4808","pdf_url":"https://www.aclweb.org/anthology/W19-4808.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://www.aclweb.org/anthology/W19-4808.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5044735229","display_name":"Jesse Vig","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jesse Vig","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5051184573","display_name":"Yonatan Belinkov","orcid":null},"institutions":[{"id":"https://openalex.org/I2801851002","display_name":"Harvard University Press","ror":"https://ror.org/006v7bf86","country_code":"US","type":"other","lineage":["https://openalex.org/I136199984","https://openalex.org/I2801851002"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yonatan Belinkov","raw_affiliation_strings":["Harvard University ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Harvard University ,","institution_ids":["https://openalex.org/I2801851002"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":24,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"63","last_page":"76"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":1.0,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":1.0,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9998999834060669,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9976999759674072,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.8591991066932678},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7375417947769165},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.6139793395996094},{"id":"https://openalex.org/keywords/syntax","display_name":"Syntax","score":0.559215784072876},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.5395521521568298},{"id":"https://openalex.org/keywords/language-understanding","display_name":"Language understanding","score":0.5311229228973389},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.49295639991760254},{"id":"https://openalex.org/keywords/dependency","display_name":"Dependency (UML)","score":0.4831208884716034},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.0928255021572113}],"concepts":[{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.8591991066932678},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7375417947769165},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.6139793395996094},{"id":"https://openalex.org/C60048249","wikidata":"https://www.wikidata.org/wiki/Q37437","display_name":"Syntax","level":2,"score":0.559215784072876},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5395521521568298},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.5311229228973389},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49295639991760254},{"id":"https://openalex.org/C19768560","wikidata":"https://www.wikidata.org/wiki/Q320727","display_name":"Dependency (UML)","level":2,"score":0.4831208884716034},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.0928255021572113},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0}],"mesh":[],"locations_count":4,"locations":[{"id":"doi:10.18653/v1/w19-4808","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/w19-4808","pdf_url":"https://www.aclweb.org/anthology/W19-4808.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:1906.04284","is_oa":true,"landing_page_url":"http://arxiv.org/abs/1906.04284","pdf_url":"https://arxiv.org/pdf/1906.04284","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},{"id":"doi:10.48550/arxiv.1906.04284","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.1906.04284","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"},{"id":"mag:2949603537","is_oa":false,"landing_page_url":null,"pdf_url":null,"source":null,"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":null}],"best_oa_location":{"id":"doi:10.18653/v1/w19-4808","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/w19-4808","pdf_url":"https://www.aclweb.org/anthology/W19-4808.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP","raw_type":"proceedings-article"},"sustainable_development_goals":[{"score":0.7200000286102295,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W2949603537.pdf","grobid_xml":"https://content.openalex.org/works/W2949603537.grobid-xml"},"referenced_works_count":27,"referenced_works":["https://openalex.org/W2118463056","https://openalex.org/W2610881169","https://openalex.org/W2626778328","https://openalex.org/W2760327630","https://openalex.org/W2799051177","https://openalex.org/W2799118206","https://openalex.org/W2890353432","https://openalex.org/W2896457183","https://openalex.org/W2896667998","https://openalex.org/W2906152891","https://openalex.org/W2910243263","https://openalex.org/W2911109671","https://openalex.org/W2934842096","https://openalex.org/W2940744433","https://openalex.org/W2945767825","https://openalex.org/W2946359678","https://openalex.org/W2946417913","https://openalex.org/W2950784811","https://openalex.org/W2951528897","https://openalex.org/W2962739339","https://openalex.org/W2962958286","https://openalex.org/W2962965405","https://openalex.org/W2963418779","https://openalex.org/W2963651521","https://openalex.org/W2964174820","https://openalex.org/W2964303116","https://openalex.org/W2964308564"],"related_works":["https://openalex.org/W2972498556","https://openalex.org/W2963341956","https://openalex.org/W2963403868","https://openalex.org/W3021530909","https://openalex.org/W2972324944","https://openalex.org/W2951025380","https://openalex.org/W2948140294","https://openalex.org/W1902237438","https://openalex.org/W3092614725","https://openalex.org/W3098839135","https://openalex.org/W2960909015","https://openalex.org/W2977436776","https://openalex.org/W2978946537","https://openalex.org/W3171900338","https://openalex.org/W2995244220","https://openalex.org/W3153369949","https://openalex.org/W2998964503","https://openalex.org/W3014866981","https://openalex.org/W2889029893","https://openalex.org/W3010763821"],"abstract_inverted_index":{"The":[0],"Transformer":[1,32],"is":[2],"a":[3,16,31,55],"fully":[4],"attention-based":[5],"alternative":[6],"to":[7],"recurrent":[8],"networks":[9],"that":[10,60,75,90,107],"has":[11],"achieved":[12],"state-of-the-art":[13],"results":[14],"across":[15],"range":[17],"of":[18,28,65,94],"NLP":[19],"tasks.":[20],"In":[21],"this":[22],"paper,":[23],"we":[24,103],"analyze":[25,47],"the":[26,35,48,72,84,91,95,98],"structure":[27],"attention":[29,42,51,61,76,115],"in":[30,83],"language":[33],"model,":[34,73],"GPT-2":[36],"small":[37],"pretrained":[38],"model.":[39],"We":[40,58,87],"visualize":[41],"for":[43],"individual":[44],"instances":[45],"and":[46,52,74],"interaction":[49],"between":[50],"syntax":[53],"over":[54],"large":[56],"corpus.":[57],"find":[59,89],"targets":[62],"different":[63,68],"parts":[64],"speech":[66],"at":[67],"layer":[69],"depths":[70],"within":[71],"aligns":[77],"with":[78],"dependency":[79],"relations":[80],"most":[81,99],"strongly":[82],"middle":[85],"layers.":[86],"also":[88],"deepest":[92],"layers":[93],"model":[96],"capture":[97],"distant":[100],"relationships.":[101],"Finally,":[102],"extract":[104],"exemplar":[105],"sentences":[106],"reveal":[108],"highly":[109],"specific":[110],"patterns":[111],"targeted":[112],"by":[113],"particular":[114],"heads.":[116]},"counts_by_year":[{"year":2025,"cited_by_count":1},{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":4},{"year":2022,"cited_by_count":5},{"year":2021,"cited_by_count":7},{"year":2020,"cited_by_count":4},{"year":2019,"cited_by_count":2}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}