The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer

Torec T. Luik; Miguel Rios; Ameen Abu-Hanna; Henk C. P. M. van Weert; Martijn C. Schut

doi:https://doi.org/10.1007/978-3-030-77211-6_51

The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer

Torec T. Luik, Miguel Rios, Ameen Abu-Hanna, Henk C. P. M. van Weert, Martijn C. Schut

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Neural models that use context-dependency in the learned text are computationally expensive. We compare the effectiveness (predictive performance) and efficiency (computational effort) of a context-independent Phrase Skip-Gram (PSG) model and a contextualized Hierarchical Attention Network (HAN) model for early prediction of lung cancer using free-text patient files from Dutch primary care physicians. The performance of PSG (AUROC 0.74 (0.69–0.79)) was comparable to HAN (AUROC 0.73 (0.68–0.78)); it achieved better calibration; had much less parameters (301 versus > 300k) and much faster (36 versus 460 s). This demonstrates an important case in which the complex contextualized neural models were not required.

Original language	English
Title of host publication	Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings
Editors	Allan Tucker, Pedro Henriques Abreu, Jaime Cardoso, Pedro Pereira Rodrigues, David Riaño
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	433-437
Number of pages	5
Volume	12721 LNAI
ISBN (Print)	9783030772109
DOIs	https://doi.org/10.1007/978-3-030-77211-6_51
Publication status	Published - 2021
Event	19th International Conference on Artiﬁcial Intelligence in Medicine, AIME 2021 - Virtual, Online Duration: 15 Jun 2021 → 18 Jun 2021

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12721 LNAI

Conference

Conference	19th International Conference on Artiﬁcial Intelligence in Medicine, AIME 2021
City	Virtual, Online
Period	15/06/2021 → 18/06/2021

Keywords

Cancer
Deep learning
N-Grams
Phrase skip-gram
Prediction models
Primary care
Word embeddings

Access to Document

https://doi.org/10.1007/978-3-030-77211-6_51

Cite this

Luik, T. T., Rios, M., Abu-Hanna, A., van Weert, H. C. P. M., & Schut, M. C. (2021). The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer. In A. Tucker, P. Henriques Abreu, J. Cardoso, P. Pereira Rodrigues, & D. Riaño (Eds.), Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings (Vol. 12721 LNAI, pp. 433-437). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12721 LNAI). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-77211-6_51

Luik, Torec T. ; Rios, Miguel ; Abu-Hanna, Ameen et al. / The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer. Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings. editor / Allan Tucker ; Pedro Henriques Abreu ; Jaime Cardoso ; Pedro Pereira Rodrigues ; David Riaño. Vol. 12721 LNAI Springer Science and Business Media Deutschland GmbH, 2021. pp. 433-437 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{03dc6d72f0b743308e8cd38baa0a7ceb,

title = "The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer",

abstract = "Neural models that use context-dependency in the learned text are computationally expensive. We compare the effectiveness (predictive performance) and efficiency (computational effort) of a context-independent Phrase Skip-Gram (PSG) model and a contextualized Hierarchical Attention Network (HAN) model for early prediction of lung cancer using free-text patient files from Dutch primary care physicians. The performance of PSG (AUROC 0.74 (0.69–0.79)) was comparable to HAN (AUROC 0.73 (0.68–0.78)); it achieved better calibration; had much less parameters (301 versus > 300k) and much faster (36 versus 460 s). This demonstrates an important case in which the complex contextualized neural models were not required.",

keywords = "Cancer, Deep learning, N-Grams, Phrase skip-gram, Prediction models, Primary care, Word embeddings",

author = "Luik, {Torec T.} and Miguel Rios and Ameen Abu-Hanna and {van Weert}, {Henk C. P. M.} and Schut, {Martijn C.}",

note = "Publisher Copyright: {\textcopyright} 2021, Springer Nature Switzerland AG. Copyright: Copyright 2021 Elsevier B.V., All rights reserved.; 19th International Conference on Artiﬁcial Intelligence in Medicine, AIME 2021 ; Conference date: 15-06-2021 Through 18-06-2021",

year = "2021",

doi = "https://doi.org/10.1007/978-3-030-77211-6_51",

language = "English",

isbn = "9783030772109",

volume = "12721 LNAI",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "433--437",

editor = "Allan Tucker and {Henriques Abreu}, Pedro and Jaime Cardoso and {Pereira Rodrigues}, Pedro and David Ria{\~n}o",

booktitle = "Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings",

address = "Germany",

}

Luik, TT, Rios, M, Abu-Hanna, A , van Weert, HCPM & Schut, MC 2021, The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer. in A Tucker, P Henriques Abreu, J Cardoso, P Pereira Rodrigues & D Riaño (eds), Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings. vol. 12721 LNAI, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12721 LNAI, Springer Science and Business Media Deutschland GmbH, pp. 433-437, 19th International Conference on Artiﬁcial Intelligence in Medicine, AIME 2021, Virtual, Online, 15/06/2021. https://doi.org/10.1007/978-3-030-77211-6_51

The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer. / Luik, Torec T.; Rios, Miguel; Abu-Hanna, Ameen et al.
Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings. ed. / Allan Tucker; Pedro Henriques Abreu; Jaime Cardoso; Pedro Pereira Rodrigues; David Riaño. Vol. 12721 LNAI Springer Science and Business Media Deutschland GmbH, 2021. p. 433-437 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12721 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer

AU - Luik, Torec T.

AU - Rios, Miguel

AU - Abu-Hanna, Ameen

AU - van Weert, Henk C. P. M.

AU - Schut, Martijn C.

PY - 2021

Y1 - 2021

N2 - Neural models that use context-dependency in the learned text are computationally expensive. We compare the effectiveness (predictive performance) and efficiency (computational effort) of a context-independent Phrase Skip-Gram (PSG) model and a contextualized Hierarchical Attention Network (HAN) model for early prediction of lung cancer using free-text patient files from Dutch primary care physicians. The performance of PSG (AUROC 0.74 (0.69–0.79)) was comparable to HAN (AUROC 0.73 (0.68–0.78)); it achieved better calibration; had much less parameters (301 versus > 300k) and much faster (36 versus 460 s). This demonstrates an important case in which the complex contextualized neural models were not required.

AB - Neural models that use context-dependency in the learned text are computationally expensive. We compare the effectiveness (predictive performance) and efficiency (computational effort) of a context-independent Phrase Skip-Gram (PSG) model and a contextualized Hierarchical Attention Network (HAN) model for early prediction of lung cancer using free-text patient files from Dutch primary care physicians. The performance of PSG (AUROC 0.74 (0.69–0.79)) was comparable to HAN (AUROC 0.73 (0.68–0.78)); it achieved better calibration; had much less parameters (301 versus > 300k) and much faster (36 versus 460 s). This demonstrates an important case in which the complex contextualized neural models were not required.

KW - Cancer

KW - Deep learning

KW - N-Grams

KW - Phrase skip-gram

KW - Prediction models

KW - Primary care

KW - Word embeddings

UR - http://www.scopus.com/inward/record.url?scp=85111380022&partnerID=8YFLogxK

U2 - https://doi.org/10.1007/978-3-030-77211-6_51

DO - https://doi.org/10.1007/978-3-030-77211-6_51

M3 - Conference contribution

SN - 9783030772109

VL - 12721 LNAI

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 433

EP - 437

BT - Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings

A2 - Tucker, Allan

A2 - Henriques Abreu, Pedro

A2 - Cardoso, Jaime

A2 - Pereira Rodrigues, Pedro

A2 - Riaño, David

PB - Springer Science and Business Media Deutschland GmbH

T2 - 19th International Conference on Artiﬁcial Intelligence in Medicine, AIME 2021

Y2 - 15 June 2021 through 18 June 2021

ER -

Luik TT, Rios M, Abu-Hanna A , van Weert HCPM , Schut MC. The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer. In Tucker A, Henriques Abreu P, Cardoso J, Pereira Rodrigues P, Riaño D, editors, Artificial Intelligence in Medicine - 19th International Conference on Artificial Intelligence in Medicine, AIME 2021, Proceedings. Vol. 12721 LNAI. Springer Science and Business Media Deutschland GmbH. 2021. p. 433-437. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: https://doi.org/10.1007/978-3-030-77211-6_51

The Effectiveness of Phrase Skip-Gram in Primary Care NLP for the Prediction of Lung Cancer

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Cite this