Predicting Annotation Difficulty to Improve Task Routing and Model Performance for Biomedical Information Extraction

Modern NLP systems require high-quality annotated data. In specialized domains, expert annotations may be prohibitively expensive. An alternative is to rely on crowdsourcing to reduce costs at the risk of introducing noise. In this paper we demonstrate that directly modeling instance difficulty can...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yang, Yinfei, Agarwal, Oshin, Tar, Chris, Wallace, Byron C, Nenkova, Ani
Format: Artikel
Sprache:eng
Schlagworte:
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 Yang, Yinfei
Agarwal, Oshin
Tar, Chris
Wallace, Byron C
Nenkova, Ani
description Modern NLP systems require high-quality annotated data. In specialized domains, expert annotations may be prohibitively expensive. An alternative is to rely on crowdsourcing to reduce costs at the risk of introducing noise. In this paper we demonstrate that directly modeling instance difficulty can be used to improve model performance, and to route instances to appropriate annotators. Our difficulty prediction model combines two learned representations: a `universal' encoder trained on out-of-domain data, and a task-specific encoder. Experiments on a complex biomedical information extraction task using expert and lay annotators show that: (i) simply excluding from the training data instances predicted to be difficult yields a small boost in performance; (ii) using difficulty scores to weight instances during training provides further, consistent gains; (iii) assigning instances predicted to be difficult to domain experts is an effective strategy for task routing. Our experiments confirm the expectation that for specialized tasks expert annotations are higher quality than crowd labels, and hence preferable to obtain if practical. Moreover, augmenting small amounts of expert data with a larger set of lay annotations leads to further improvements in model performance.
doi_str_mv 10.48550/arxiv.1905.07791
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_1905_07791</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>1905_07791</sourcerecordid><originalsourceid>FETCH-LOGICAL-a671-3394ec1f4487fb1940b1f1f81b110a7b662a6a8199da775981d49b2746741a393</originalsourceid><addsrcrecordid>eNotj8FOwzAQRH3hgAofwIn9gQRv48TxsZQCkYqoUO7RJrGRRWJXrlu1fw8JPc1IM3rSY-wBeSrKPOdPFM72lKLiecqlVHjL4i7o3nbRum9YOecjResdvFhjbHcc4gWih2rcB3_SUNPhB778cX6T6-HD93qAnQ7Gh5Fcp-GvwLP14wSlASo3LzNzc46BuqnesRtDw0HfX3PB6tdNvX5Ptp9v1Xq1TaiQmGSZErpDI0QpTYtK8BYNmhJbRE6yLYolFVSiUj1JmasSe6HapRSFFEiZyhbs8R87azf7YEcKl2bSb2b97Be38lZ4</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Predicting Annotation Difficulty to Improve Task Routing and Model Performance for Biomedical Information Extraction</title><source>arXiv.org</source><creator>Yang, Yinfei ; Agarwal, Oshin ; Tar, Chris ; Wallace, Byron C ; Nenkova, Ani</creator><creatorcontrib>Yang, Yinfei ; Agarwal, Oshin ; Tar, Chris ; Wallace, Byron C ; Nenkova, Ani</creatorcontrib><description>Modern NLP systems require high-quality annotated data. In specialized domains, expert annotations may be prohibitively expensive. An alternative is to rely on crowdsourcing to reduce costs at the risk of introducing noise. In this paper we demonstrate that directly modeling instance difficulty can be used to improve model performance, and to route instances to appropriate annotators. Our difficulty prediction model combines two learned representations: a `universal' encoder trained on out-of-domain data, and a task-specific encoder. Experiments on a complex biomedical information extraction task using expert and lay annotators show that: (i) simply excluding from the training data instances predicted to be difficult yields a small boost in performance; (ii) using difficulty scores to weight instances during training provides further, consistent gains; (iii) assigning instances predicted to be difficult to domain experts is an effective strategy for task routing. Our experiments confirm the expectation that for specialized tasks expert annotations are higher quality than crowd labels, and hence preferable to obtain if practical. Moreover, augmenting small amounts of expert data with a larger set of lay annotations leads to further improvements in model performance.</description><identifier>DOI: 10.48550/arxiv.1905.07791</identifier><language>eng</language><subject>Computer Science - Computation and Language</subject><creationdate>2019-05</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/1905.07791$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.1905.07791$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Yang, Yinfei</creatorcontrib><creatorcontrib>Agarwal, Oshin</creatorcontrib><creatorcontrib>Tar, Chris</creatorcontrib><creatorcontrib>Wallace, Byron C</creatorcontrib><creatorcontrib>Nenkova, Ani</creatorcontrib><title>Predicting Annotation Difficulty to Improve Task Routing and Model Performance for Biomedical Information Extraction</title><description>Modern NLP systems require high-quality annotated data. In specialized domains, expert annotations may be prohibitively expensive. An alternative is to rely on crowdsourcing to reduce costs at the risk of introducing noise. In this paper we demonstrate that directly modeling instance difficulty can be used to improve model performance, and to route instances to appropriate annotators. Our difficulty prediction model combines two learned representations: a `universal' encoder trained on out-of-domain data, and a task-specific encoder. Experiments on a complex biomedical information extraction task using expert and lay annotators show that: (i) simply excluding from the training data instances predicted to be difficult yields a small boost in performance; (ii) using difficulty scores to weight instances during training provides further, consistent gains; (iii) assigning instances predicted to be difficult to domain experts is an effective strategy for task routing. Our experiments confirm the expectation that for specialized tasks expert annotations are higher quality than crowd labels, and hence preferable to obtain if practical. Moreover, augmenting small amounts of expert data with a larger set of lay annotations leads to further improvements in model performance.</description><subject>Computer Science - Computation and Language</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2019</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8FOwzAQRH3hgAofwIn9gQRv48TxsZQCkYqoUO7RJrGRRWJXrlu1fw8JPc1IM3rSY-wBeSrKPOdPFM72lKLiecqlVHjL4i7o3nbRum9YOecjResdvFhjbHcc4gWih2rcB3_SUNPhB778cX6T6-HD93qAnQ7Gh5Fcp-GvwLP14wSlASo3LzNzc46BuqnesRtDw0HfX3PB6tdNvX5Ptp9v1Xq1TaiQmGSZErpDI0QpTYtK8BYNmhJbRE6yLYolFVSiUj1JmasSe6HapRSFFEiZyhbs8R87azf7YEcKl2bSb2b97Be38lZ4</recordid><startdate>20190519</startdate><enddate>20190519</enddate><creator>Yang, Yinfei</creator><creator>Agarwal, Oshin</creator><creator>Tar, Chris</creator><creator>Wallace, Byron C</creator><creator>Nenkova, Ani</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20190519</creationdate><title>Predicting Annotation Difficulty to Improve Task Routing and Model Performance for Biomedical Information Extraction</title><author>Yang, Yinfei ; Agarwal, Oshin ; Tar, Chris ; Wallace, Byron C ; Nenkova, Ani</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a671-3394ec1f4487fb1940b1f1f81b110a7b662a6a8199da775981d49b2746741a393</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2019</creationdate><topic>Computer Science - Computation and Language</topic><toplevel>online_resources</toplevel><creatorcontrib>Yang, Yinfei</creatorcontrib><creatorcontrib>Agarwal, Oshin</creatorcontrib><creatorcontrib>Tar, Chris</creatorcontrib><creatorcontrib>Wallace, Byron C</creatorcontrib><creatorcontrib>Nenkova, Ani</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Yang, Yinfei</au><au>Agarwal, Oshin</au><au>Tar, Chris</au><au>Wallace, Byron C</au><au>Nenkova, Ani</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Predicting Annotation Difficulty to Improve Task Routing and Model Performance for Biomedical Information Extraction</atitle><date>2019-05-19</date><risdate>2019</risdate><abstract>Modern NLP systems require high-quality annotated data. In specialized domains, expert annotations may be prohibitively expensive. An alternative is to rely on crowdsourcing to reduce costs at the risk of introducing noise. In this paper we demonstrate that directly modeling instance difficulty can be used to improve model performance, and to route instances to appropriate annotators. Our difficulty prediction model combines two learned representations: a `universal' encoder trained on out-of-domain data, and a task-specific encoder. Experiments on a complex biomedical information extraction task using expert and lay annotators show that: (i) simply excluding from the training data instances predicted to be difficult yields a small boost in performance; (ii) using difficulty scores to weight instances during training provides further, consistent gains; (iii) assigning instances predicted to be difficult to domain experts is an effective strategy for task routing. Our experiments confirm the expectation that for specialized tasks expert annotations are higher quality than crowd labels, and hence preferable to obtain if practical. Moreover, augmenting small amounts of expert data with a larger set of lay annotations leads to further improvements in model performance.</abstract><doi>10.48550/arxiv.1905.07791</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.1905.07791
ispartof
issn
language eng
recordid cdi_arxiv_primary_1905_07791
source arXiv.org
subjects Computer Science - Computation and Language
title Predicting Annotation Difficulty to Improve Task Routing and Model Performance for Biomedical Information Extraction
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-24T01%3A40%3A33IST&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=Predicting%20Annotation%20Difficulty%20to%20Improve%20Task%20Routing%20and%20Model%20Performance%20for%20Biomedical%20Information%20Extraction&rft.au=Yang,%20Yinfei&rft.date=2019-05-19&rft_id=info:doi/10.48550/arxiv.1905.07791&rft_dat=%3Carxiv_GOX%3E1905_07791%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