{"id":"https://openalex.org/W4281550655","doi":"https://doi.org/10.48550/arxiv.2205.12228","title":"When More Data Hurts: A Troubling Quirk in Developing Broad-Coverage Natural Language Understanding Systems","display_name":"When More Data Hurts: A Troubling Quirk in Developing Broad-Coverage Natural Language Understanding Systems","publication_year":2022,"publication_date":"2022-05-24","ids":{"openalex":"https://openalex.org/W4281550655","doi":"https://doi.org/10.48550/arxiv.2205.12228"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2205.12228","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2205.12228","pdf_url":"https://arxiv.org/pdf/2205.12228","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2205.12228","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5038276570","display_name":"Elias Stengel-Eskin","orcid":"https://orcid.org/0000-0002-6689-505X","observed_orcids":["https://orcid.org/0000-0002-6689-505X"]},"institutions":[{"id":"https://openalex.org/I145311948","display_name":"Johns Hopkins University","ror":"https://ror.org/00za53h95","country_code":"US","type":"education","lineage":["https://openalex.org/I145311948"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Stengel-Eskin, Elias","raw_affiliation_strings":["Johns Hopkins University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Johns Hopkins University","institution_ids":["https://openalex.org/I145311948"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5032039796","display_name":"Emmanouil Antonios Platanios","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Platanios, Emmanouil Antonios","raw_affiliation_strings":["Microsoft Semantic Machines"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Semantic Machines","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001722321","display_name":"Adam Pauls","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pauls, Adam","raw_affiliation_strings":["Microsoft Semantic Machines"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Semantic Machines","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048087523","display_name":"Sam Thomson","orcid":"https://orcid.org/0009-0004-3099-5050","observed_orcids":["https://orcid.org/0009-0004-3099-5050"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thomson, Sam","raw_affiliation_strings":["Microsoft Semantic Machines"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Semantic Machines","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100604056","display_name":"Hao Fang","orcid":"https://orcid.org/0000-0001-7194-5771","observed_orcids":["https://orcid.org/0000-0001-7194-5771","https://orcid.org/0009-0004-0271-6579"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fang, Hao","raw_affiliation_strings":["Microsoft Semantic Machines"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Semantic Machines","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075825791","display_name":"Benjamin Van Durme","orcid":"https://orcid.org/0000-0003-4328-4288","observed_orcids":["https://orcid.org/0000-0003-4328-4288"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Van Durme, Benjamin","raw_affiliation_strings":["Microsoft Semantic Machines"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Semantic Machines","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052467896","display_name":"Jason M. Eisner","orcid":"https://orcid.org/0000-0002-8861-0772","observed_orcids":["https://orcid.org/0000-0002-8861-0772"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Eisner, Jason","raw_affiliation_strings":["Microsoft Semantic Machines"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Semantic Machines","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5075435632","display_name":"Yu Su","orcid":"https://orcid.org/0000-0002-6649-4766","observed_orcids":["https://orcid.org/0000-0002-6649-4766","https://orcid.org/0000-0003-4306-0161"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Su, Yu","raw_affiliation_strings":["Microsoft Semantic Machines"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Semantic Machines","institution_ids":[]}]}],"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":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.42480000853538513,"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":0.42480000853538513,"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":0.42260000109672546,"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/T12031","display_name":"Speech and dialogue systems","score":0.11219999939203262,"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/natural-language-understanding","display_name":"natural language understanding","score":0.8140000104904175},{"id":"https://openalex.org/keywords/training-set-size","display_name":"training data size","score":0.746999979019165},{"id":"https://openalex.org/keywords/lexical-cues","display_name":"lexical cues","score":0.8309999704360962},{"id":"https://openalex.org/keywords/intent-recognition","display_name":"intent classification","score":0.8130000233650208},{"id":"https://openalex.org/keywords/semantic-parsing","display_name":"semantic parsing","score":1.0},{"id":"https://openalex.org/keywords/class-imbalance","display_name":"class imbalance","score":0.8130000233650208},{"id":"https://openalex.org/keywords/training-sample-selection","display_name":"training data selection","score":0.4480000138282776}],"concepts":[{"id":"https://openalex.org/C134400042","wikidata":"https://www.wikidata.org/wiki/Q2372244","display_name":"Symbol (formal)","level":2,"score":0.6943097114562988},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6767521500587463},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.6210876703262329},{"id":"https://openalex.org/C186644900","wikidata":"https://www.wikidata.org/wiki/Q194152","display_name":"Parsing","level":2,"score":0.6147269606590271},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5727337598800659},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.5589213967323303},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5053616166114807},{"id":"https://openalex.org/C2780876879","wikidata":"https://www.wikidata.org/wiki/Q3054749","display_name":"Meaning (existential)","level":2,"score":0.4950694143772125},{"id":"https://openalex.org/C554936623","wikidata":"https://www.wikidata.org/wiki/Q199657","display_name":"Reading (process)","level":2,"score":0.44905662536621094},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.367246150970459},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.32018202543258667},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.19793394207954407},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.1247352659702301},{"id":"https://openalex.org/C542102704","wikidata":"https://www.wikidata.org/wiki/Q183257","display_name":"Psychotherapist","level":1,"score":0.0},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C94625758","wikidata":"https://www.wikidata.org/wiki/Q7163","display_name":"Politics","level":2,"score":0.0},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.0},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2205.12228","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2205.12228","pdf_url":"https://arxiv.org/pdf/2205.12228","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},{"id":"doi:10.48550/arxiv.2205.12228","doi":"https://doi.org/10.48550/arxiv.2205.12228","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2205.12228","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2205.12228","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2205.12228","pdf_url":"https://arxiv.org/pdf/2205.12228","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[{"score":0.7799999713897705,"id":"https://metadata.un.org/sdg/4","display_name":"Quality education"}],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2359896531","https://openalex.org/W2293063786","https://openalex.org/W2911292476","https://openalex.org/W2367925007","https://openalex.org/W4288263119","https://openalex.org/W3015724364","https://openalex.org/W2967994095","https://openalex.org/W4285240985","https://openalex.org/W2900126711","https://openalex.org/W4286930972"],"abstract_inverted_index":{"In":[0],"natural":[1],"language":[2],"understanding":[3],"(NLU)":[4],"production":[5],"systems,":[6],"users'":[7],"evolving":[8],"needs":[9],"necessitate":[10],"the":[11,24,40,62,68,122,128,146,152,165,168],"addition":[12],"of":[13,44,170],"new":[14,20,69,91,147],"features":[15],"over":[16],"time,":[17],"indexed":[18],"by":[19],"symbols":[21,92],"added":[22],"to":[23,89,160],"meaning":[25],"representation":[26],"space.":[27],"This":[28,82],"requires":[29],"additional":[30],"training":[31,63,80,96,153,158],"data":[32,182],"and":[33,115,181],"results":[34],"in":[35,56],"ever-growing":[36],"datasets.":[37],"We":[38,98],"present":[39],"first":[41],"systematic":[42],"investigation":[43],"this":[45,101],"incremental":[46],"symbol":[47,70,148],"learning":[48],"scenario.":[49],"Our":[50],"analysis":[51],"reveals":[52],"a":[53,94],"troubling":[54],"quirk":[55],"building":[57],"broad-coverage":[58],"NLU":[59,111,173],"systems:":[60],"as":[61,121,151],"dataset":[64,154],"grows,":[65],"performance":[66],"on":[67,108,175],"often":[71,163],"decreases":[72],"if":[73],"we":[74,125,136],"do":[75],"not":[76],"accordingly":[77],"increase":[78],"its":[79],"data.":[81],"suggests":[83],"that":[84,100,127],"it":[85],"becomes":[86],"more":[87],"difficult":[88],"learn":[90],"with":[93,133],"larger":[95],"dataset.":[97],"show":[99],"trend":[102,129],"holds":[103],"for":[104,145],"multiple":[105],"mainstream":[106,171],"models":[107,174],"two":[109],"common":[110],"tasks:":[112],"intent":[113],"recognition":[114],"semantic":[116],"parsing.":[117],"Rejecting":[118],"class":[119],"imbalance":[120],"sole":[123],"culprit,":[124],"reveal":[126],"is":[130],"closely":[131],"associated":[132],"an":[134],"effect":[135],"call":[137],"source":[138],"signal":[139],"dilution,":[140],"where":[141],"strong":[142],"lexical":[143,177],"cues":[144],"become":[149],"diluted":[150],"grows.":[155],"Selectively":[156],"dropping":[157],"examples":[159],"prevent":[161],"dilution":[162],"reverses":[164],"trend,":[166],"showing":[167],"over-reliance":[169],"neural":[172],"simple":[176],"cues.":[178],"Code,":[179],"models,":[180],"are":[183],"available":[184],"at":[185],"https://aka.ms/nlu-incremental-symbol-learning":[186]},"counts_by_year":[],"updated_date":"2026-10-13T10:08:12.033832","created_date":"2025-10-10T00:00:00"}