Error Detection in Large-Scale Natural Language Understanding Systems Using Transformer Models

Error Detection in Large-Scale Natural Language Understanding Systems Using Transformer Models

Large-scale conversational assistants like Alexa, Siri, Cortana and Google Assistant process every utterance using multiple models for domain, intent and named entity recognition. Given the decoupled nature of model development and large traffic volumes, it is extremely difficult to identify utteran...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Chada, Rakesh, Natarajan, Pradeep, Fofadiya, Darshan, Ramachandra, Prathap
Format: Artikel
Sprache:eng
Schlagworte:
Computer Science - Artificial Intelligence
Computer Science - Computation and Language
Computer Science - Learning
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Chada, Rakesh
Natarajan, Pradeep
Fofadiya, Darshan
Ramachandra, Prathap
description Large-scale conversational assistants like Alexa, Siri, Cortana and Google Assistant process every utterance using multiple models for domain, intent and named entity recognition. Given the decoupled nature of model development and large traffic volumes, it is extremely difficult to identify utterances processed erroneously by such systems. We address this challenge to detect domain classification errors using offline Transformer models. We combine utterance encodings from a RoBERTa model with the Nbest hypothesis produced by the production system. We then fine-tune end-to-end in a multitask setting using a small dataset of humanannotated utterances with domain classification errors. We tested our approach for detecting misclassifications from one domain that accounts for
doi_str_mv 10.48550/arxiv.2109.01754
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2109_01754</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2109_01754</sourcerecordid><originalsourceid>FETCH-LOGICAL-a674-dfa86b2f15d7e93936ed594142dfad361600973373cd55062d50b91138b54d4a3</originalsourceid><addsrcrecordid>eNotz8tOwzAQBVBvWKDCB7DCP5Bgx4_ES1TKQ0ph0XRLNIknUaTEQeMU0b-nLaxGd650pcPYnRSpLowRD0A_w3eaSeFSIXOjr9nnhmgm_oQLtsswBz4EXgL1mOxaGJG_w3IgGE-_0B-gR74PHikuEPwQer47xgWnyPfxnCqCELuZJiS-nT2O8YZddTBGvP2_K1Y9b6r1a1J-vLytH8sEbK4T30Fhm6yTxufolFMWvXFa6uzUeGWlFcLlSuWq9SeGzbwRjZNSFY3RXoNasfu_2Quw_qJhAjrWZ2h9gapfF9pOKQ</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Error Detection in Large-Scale Natural Language Understanding Systems Using Transformer Models</title><source>arXiv.org</source><creator>Chada, Rakesh ; Natarajan, Pradeep ; Fofadiya, Darshan ; Ramachandra, Prathap</creator><creatorcontrib>Chada, Rakesh ; Natarajan, Pradeep ; Fofadiya, Darshan ; Ramachandra, Prathap</creatorcontrib><description>Large-scale conversational assistants like Alexa, Siri, Cortana and Google Assistant process every utterance using multiple models for domain, intent and named entity recognition. Given the decoupled nature of model development and large traffic volumes, it is extremely difficult to identify utterances processed erroneously by such systems. We address this challenge to detect domain classification errors using offline Transformer models. We combine utterance encodings from a RoBERTa model with the Nbest hypothesis produced by the production system. We then fine-tune end-to-end in a multitask setting using a small dataset of humanannotated utterances with domain classification errors. We tested our approach for detecting misclassifications from one domain that accounts for &lt;0.5% of the traffic in a large-scale conversational AI system. Our approach achieves an F1 score of 30% outperforming a bi- LSTM baseline by 16.9% and a standalone RoBERTa model by 4.8%. We improve this further by 2.2% to 32.2% by ensembling multiple models.</description><identifier>DOI: 10.48550/arxiv.2109.01754</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Computation and Language ; Computer Science - Learning</subject><creationdate>2021-09</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2109.01754$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2109.01754$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Chada, Rakesh</creatorcontrib><creatorcontrib>Natarajan, Pradeep</creatorcontrib><creatorcontrib>Fofadiya, Darshan</creatorcontrib><creatorcontrib>Ramachandra, Prathap</creatorcontrib><title>Error Detection in Large-Scale Natural Language Understanding Systems Using Transformer Models</title><description>Large-scale conversational assistants like Alexa, Siri, Cortana and Google Assistant process every utterance using multiple models for domain, intent and named entity recognition. Given the decoupled nature of model development and large traffic volumes, it is extremely difficult to identify utterances processed erroneously by such systems. We address this challenge to detect domain classification errors using offline Transformer models. We combine utterance encodings from a RoBERTa model with the Nbest hypothesis produced by the production system. We then fine-tune end-to-end in a multitask setting using a small dataset of humanannotated utterances with domain classification errors. We tested our approach for detecting misclassifications from one domain that accounts for &lt;0.5% of the traffic in a large-scale conversational AI system. Our approach achieves an F1 score of 30% outperforming a bi- LSTM baseline by 16.9% and a standalone RoBERTa model by 4.8%. We improve this further by 2.2% to 32.2% by ensembling multiple models.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Computation and Language</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz8tOwzAQBVBvWKDCB7DCP5Bgx4_ES1TKQ0ph0XRLNIknUaTEQeMU0b-nLaxGd650pcPYnRSpLowRD0A_w3eaSeFSIXOjr9nnhmgm_oQLtsswBz4EXgL1mOxaGJG_w3IgGE-_0B-gR74PHikuEPwQer47xgWnyPfxnCqCELuZJiS-nT2O8YZddTBGvP2_K1Y9b6r1a1J-vLytH8sEbK4T30Fhm6yTxufolFMWvXFa6uzUeGWlFcLlSuWq9SeGzbwRjZNSFY3RXoNasfu_2Quw_qJhAjrWZ2h9gapfF9pOKQ</recordid><startdate>20210903</startdate><enddate>20210903</enddate><creator>Chada, Rakesh</creator><creator>Natarajan, Pradeep</creator><creator>Fofadiya, Darshan</creator><creator>Ramachandra, Prathap</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20210903</creationdate><title>Error Detection in Large-Scale Natural Language Understanding Systems Using Transformer Models</title><author>Chada, Rakesh ; Natarajan, Pradeep ; Fofadiya, Darshan ; Ramachandra, Prathap</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a674-dfa86b2f15d7e93936ed594142dfad361600973373cd55062d50b91138b54d4a3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Computation and Language</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Chada, Rakesh</creatorcontrib><creatorcontrib>Natarajan, Pradeep</creatorcontrib><creatorcontrib>Fofadiya, Darshan</creatorcontrib><creatorcontrib>Ramachandra, Prathap</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Chada, Rakesh</au><au>Natarajan, Pradeep</au><au>Fofadiya, Darshan</au><au>Ramachandra, Prathap</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Error Detection in Large-Scale Natural Language Understanding Systems Using Transformer Models</atitle><date>2021-09-03</date><risdate>2021</risdate><abstract>Large-scale conversational assistants like Alexa, Siri, Cortana and Google Assistant process every utterance using multiple models for domain, intent and named entity recognition. Given the decoupled nature of model development and large traffic volumes, it is extremely difficult to identify utterances processed erroneously by such systems. We address this challenge to detect domain classification errors using offline Transformer models. We combine utterance encodings from a RoBERTa model with the Nbest hypothesis produced by the production system. We then fine-tune end-to-end in a multitask setting using a small dataset of humanannotated utterances with domain classification errors. We tested our approach for detecting misclassifications from one domain that accounts for &lt;0.5% of the traffic in a large-scale conversational AI system. Our approach achieves an F1 score of 30% outperforming a bi- LSTM baseline by 16.9% and a standalone RoBERTa model by 4.8%. We improve this further by 2.2% to 32.2% by ensembling multiple models.</abstract><doi>10.48550/arxiv.2109.01754</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2109.01754
ispartof
issn
language eng
recordid cdi_arxiv_primary_2109_01754
source arXiv.org
subjects Computer Science - Artificial Intelligence
Computer Science - Computation and Language
Computer Science - Learning
title Error Detection in Large-Scale Natural Language Understanding Systems Using Transformer Models
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T20%3A15%3A54IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Error%20Detection%20in%20Large-Scale%20Natural%20Language%20Understanding%20Systems%20Using%20Transformer%20Models&rft.au=Chada,%20Rakesh&rft.date=2021-09-03&rft_id=info:doi/10.48550/arxiv.2109.01754&rft_dat=%3Carxiv_GOX%3E2109_01754%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true