{"id":"https://openalex.org/W4392688329","doi":"https://doi.org/10.1145/3589335.3651526","title":"Is Cosine-Similarity of Embeddings Really About Similarity?","display_name":"Is Cosine-Similarity of Embeddings Really About Similarity?","publication_year":2024,"publication_date":"2024-05-12","ids":{"openalex":"https://openalex.org/W4392688329","doi":"https://doi.org/10.1145/3589335.3651526"},"language":"en","primary_location":{"id":"doi:10.1145/3589335.3651526","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3589335.3651526","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3589335.3651526","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Companion Proceedings of the ACM Web Conference 2024","raw_type":"proceedings-article"},"type":"preprint","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3589335.3651526","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5070327072","display_name":"Harald Steck","orcid":"https://orcid.org/0009-0007-7448-8335"},"institutions":[{"id":"https://openalex.org/I869089601","display_name":"Netflix (United States)","ror":"https://ror.org/0197qw696","country_code":"US","type":"company","lineage":["https://openalex.org/I869089601"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Harald Steck","raw_affiliation_strings":["Netflix Inc., Los Gatos, CA, USA"],"affiliations":[{"raw_affiliation_string":"Netflix Inc., Los Gatos, CA, USA","institution_ids":["https://openalex.org/I869089601"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079641101","display_name":"Chaitanya Ekanadham","orcid":"https://orcid.org/0009-0006-8029-5284"},"institutions":[{"id":"https://openalex.org/I869089601","display_name":"Netflix (United States)","ror":"https://ror.org/0197qw696","country_code":"US","type":"company","lineage":["https://openalex.org/I869089601"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chaitanya Ekanadham","raw_affiliation_strings":["Netflix Inc., Los Angeles, CA, USA"],"affiliations":[{"raw_affiliation_string":"Netflix Inc., Los Angeles, CA, USA","institution_ids":["https://openalex.org/I869089601"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5036921114","display_name":"Nathan Kallus","orcid":"https://orcid.org/0000-0003-1672-0507"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]},{"id":"https://openalex.org/I869089601","display_name":"Netflix (United States)","ror":"https://ror.org/0197qw696","country_code":"US","type":"company","lineage":["https://openalex.org/I869089601"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nathan Kallus","raw_affiliation_strings":["Netflix Inc. & Cornell University, New York, NY, USA"],"affiliations":[{"raw_affiliation_string":"Netflix Inc. & Cornell University, New York, NY, USA","institution_ids":["https://openalex.org/I869089601","https://openalex.org/I205783295"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5070327072"],"corresponding_institution_ids":["https://openalex.org/I869089601"],"apc_list":null,"apc_paid":null,"fwci":37.3524,"has_fulltext":true,"cited_by_count":107,"citation_normalized_percentile":{"value":0.99846207,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":99,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"887","last_page":"890"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9969000220298767,"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":0.9969000220298767,"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/T13083","display_name":"Advanced Text Analysis Techniques","score":0.9937999844551086,"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":0.9550999999046326,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/similarity","display_name":"Similarity (geometry)","score":0.7892581224441528},{"id":"https://openalex.org/keywords/cosine-similarity","display_name":"Cosine similarity","score":0.7496677041053772},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5669867396354675},{"id":"https://openalex.org/keywords/trigonometric-functions","display_name":"Trigonometric functions","score":0.4916844964027405},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3970031440258026},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.2887378931045532},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.24273273348808289}],"concepts":[{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.7892581224441528},{"id":"https://openalex.org/C2780762811","wikidata":"https://www.wikidata.org/wiki/Q1784941","display_name":"Cosine similarity","level":3,"score":0.7496677041053772},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5669867396354675},{"id":"https://openalex.org/C178009071","wikidata":"https://www.wikidata.org/wiki/Q93344","display_name":"Trigonometric functions","level":2,"score":0.4916844964027405},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3970031440258026},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2887378931045532},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.24273273348808289},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.0},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3589335.3651526","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3589335.3651526","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3589335.3651526","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Companion Proceedings of the ACM Web Conference 2024","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2403.05440","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.05440","pdf_url":"https://arxiv.org/pdf/2403.05440","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":null,"raw_type":"text"}],"best_oa_location":{"id":"doi:10.1145/3589335.3651526","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3589335.3651526","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3589335.3651526","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Companion Proceedings of the ACM Web Conference 2024","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4392688329.pdf"},"referenced_works_count":3,"referenced_works":["https://openalex.org/W2541443726","https://openalex.org/W3021397474","https://openalex.org/W3099700870"],"related_works":["https://openalex.org/W2389818373","https://openalex.org/W126212742","https://openalex.org/W2220831889","https://openalex.org/W1948687848","https://openalex.org/W2953417386","https://openalex.org/W4312683641","https://openalex.org/W3027421045","https://openalex.org/W2576320324","https://openalex.org/W2980386803","https://openalex.org/W3215994059"],"abstract_inverted_index":{"Cosine-similarity":[0],"is":[1,21],"the":[2,5,12,47,92,106,135],"cosine":[3],"of":[4,116,134],"angle":[6],"between":[7,15,26,50],"two":[8],"vectors,":[9],"or":[10],"equivalently":[11],"dot":[13],"product":[14],"their":[16],"normalizations.":[17],"A":[18],"popular":[19],"application":[20],"to":[22,32],"quantify":[23],"semantic":[24],"similarity":[25],"high-dimensional":[27],"objects":[28],"by":[29,105],"applying":[30],"cosine-similarity":[31,80,153],"a":[33,114],"learned":[34],"low-dimensional":[35],"feature":[36],"embedding.":[37],"This":[38],"can":[39,81],"work":[40],"better":[41],"but":[42],"sometimes":[43],"also":[44],"worse":[45],"than":[46],"unnormalized":[48],"dot-product":[49],"embedded":[51],"vectors":[52],"in":[53],"practice.":[54],"To":[55],"gain":[56],"insight":[57],"into":[58],"this":[59],"empirical":[60],"observation,":[61],"we":[62,148],"study":[63],"embeddings":[64],"derived":[65],"from":[66],"regularized":[67],"linear":[68,90,112],"models,":[69],"where":[70],"closed-form":[71],"solutions":[72],"facilitate":[73],"analytical":[74],"insights.":[75],"We":[76,108],"derive":[77],"analytically":[78],"how":[79],"yield":[82],"arbitrary":[83],"and":[84,128,141,154],"therefore":[85],"meaningless":[86],"'similarities.'":[87],"For":[88],"some":[89],"models":[91],"similarities":[93],"are":[94,102,119],"not":[95],"even":[96],"unique,":[97],"while":[98],"for":[99],"others":[100],"they":[101],"implicitly":[103],"controlled":[104],"regularization.":[107],"discuss":[109],"implications":[110],"beyond":[111],"models:":[113],"combination":[115],"different":[117],"regularizations":[118],"employed":[120],"when":[121,131],"learning":[122],"deep":[123],"models;":[124],"these":[125,146],"have":[126],"implicit":[127],"unintended":[129],"effects":[130],"taking":[132],"cosine-similarities":[133],"resulting":[136],"embeddings,":[137],"rendering":[138],"results":[139],"opaque":[140],"possibly":[142],"arbitrary.":[143],"Based":[144],"on":[145],"insights,":[147],"caution":[149],"against":[150],"blindly":[151],"using":[152],"outline":[155],"alternatives.":[156]},"counts_by_year":[{"year":2026,"cited_by_count":11},{"year":2025,"cited_by_count":74},{"year":2024,"cited_by_count":22}],"updated_date":"2026-03-27T14:29:43.386196","created_date":"2025-10-10T00:00:00"}