Query Expansion with Automatically Predicted Diagnosis: iRiS at TREC CDS track 2016

Danchen Zhang; Daqing He; Sanqiang Zhao; Lei Li

Query Expansion with Automatically Predicted Diagnosis: iRiS at TREC CDS track 2016

Danchen Zhang, Daqing He, Sanqiang Zhao, Lei Li

University of Pittsburgh

Research output: Contribution to conference › Paper › peer-review

2 Scopus citations

Abstract

This paper describes the participation of the iRiS team from University of Pittsburgh in the TREC Clinical Decision Support (CDS) track in 2016. According to the track requirements, 1,000 most relevant biomedical articles from the PubMed Collection were retrieved based on information needs of 30 patients with their electronic health records (EHR) notes. Our approach focuses on using MetaMap to extract medical concepts, and using Wikipedia knowledge base to predict the patient diagnosis. Consequently, the original query is expanded with the predicted diagnosis before sent to search PubMed articles. Parameters were tuned based on CDS 2014 and 2015, and Indri is used to construct the index of the collection. Our automatic runs on description ranks 2^nd and our manual runs on notes ranks 3^rd in all submitted runs.

Original language	English
State	Published - 2016
Externally published	Yes
Event	25th Text REtrieval Conference, TREC 2016 - Gaithersburg, United States Duration: 15 Nov 2016 → 18 Nov 2016

Conference

Conference	25th Text REtrieval Conference, TREC 2016
Country/Territory	United States
City	Gaithersburg
Period	15/11/16 → 18/11/16

Keywords

diagnose prediction
Medical text retrieval
query expansion

Cite this

@conference{e56de4f6eac44871aa2118048bff536a,

title = "Query Expansion with Automatically Predicted Diagnosis: iRiS at TREC CDS track 2016",

abstract = "This paper describes the participation of the iRiS team from University of Pittsburgh in the TREC Clinical Decision Support (CDS) track in 2016. According to the track requirements, 1,000 most relevant biomedical articles from the PubMed Collection were retrieved based on information needs of 30 patients with their electronic health records (EHR) notes. Our approach focuses on using MetaMap to extract medical concepts, and using Wikipedia knowledge base to predict the patient diagnosis. Consequently, the original query is expanded with the predicted diagnosis before sent to search PubMed articles. Parameters were tuned based on CDS 2014 and 2015, and Indri is used to construct the index of the collection. Our automatic runs on description ranks 2nd and our manual runs on notes ranks 3rd in all submitted runs.",

keywords = "diagnose prediction, Medical text retrieval, query expansion",

author = "Danchen Zhang and Daqing He and Sanqiang Zhao and Lei Li",

note = "Publisher Copyright: {\textcopyright} 2016 25th Text REtrieval Conference, TREC 2016 - Proceedings. All Rights Reserved.; 25th Text REtrieval Conference, TREC 2016 ; Conference date: 15-11-2016 Through 18-11-2016",

year = "2016",

language = "英语",

}

TY - CONF

T1 - Query Expansion with Automatically Predicted Diagnosis

T2 - 25th Text REtrieval Conference, TREC 2016

AU - Zhang, Danchen

AU - He, Daqing

AU - Zhao, Sanqiang

AU - Li, Lei

PY - 2016

Y1 - 2016

N2 - This paper describes the participation of the iRiS team from University of Pittsburgh in the TREC Clinical Decision Support (CDS) track in 2016. According to the track requirements, 1,000 most relevant biomedical articles from the PubMed Collection were retrieved based on information needs of 30 patients with their electronic health records (EHR) notes. Our approach focuses on using MetaMap to extract medical concepts, and using Wikipedia knowledge base to predict the patient diagnosis. Consequently, the original query is expanded with the predicted diagnosis before sent to search PubMed articles. Parameters were tuned based on CDS 2014 and 2015, and Indri is used to construct the index of the collection. Our automatic runs on description ranks 2nd and our manual runs on notes ranks 3rd in all submitted runs.

AB - This paper describes the participation of the iRiS team from University of Pittsburgh in the TREC Clinical Decision Support (CDS) track in 2016. According to the track requirements, 1,000 most relevant biomedical articles from the PubMed Collection were retrieved based on information needs of 30 patients with their electronic health records (EHR) notes. Our approach focuses on using MetaMap to extract medical concepts, and using Wikipedia knowledge base to predict the patient diagnosis. Consequently, the original query is expanded with the predicted diagnosis before sent to search PubMed articles. Parameters were tuned based on CDS 2014 and 2015, and Indri is used to construct the index of the collection. Our automatic runs on description ranks 2nd and our manual runs on notes ranks 3rd in all submitted runs.

KW - diagnose prediction

KW - Medical text retrieval

KW - query expansion

UR - http://www.scopus.com/inward/record.url?scp=85061929490&partnerID=8YFLogxK

M3 - 论文

AN - SCOPUS:85061929490

Y2 - 15 November 2016 through 18 November 2016

ER -

Query Expansion with Automatically Predicted Diagnosis: iRiS at TREC CDS track 2016

Abstract

Conference

Keywords

Other files and links

Fingerprint

Cite this