TY - JOUR

T1 - Large expert-curated database for benchmarking document similarity detection in biomedical literature search

AU - RELISH Consortium

AU - Brown, Peter

AU - Tan, Aik-Choon

AU - El-Esawi, Mohamed A.

AU - Liehr, Thomas

AU - Blanck, Oliver

AU - Gladue, Douglas P.

AU - Almeida, Gabriel M. F.

AU - Cernava, Tomislav

AU - Sorzano, Carlos O.

AU - Yeung, Andy W. K.

AU - Engel, Michael S.

AU - Chandrasekaran, Arun Richard

AU - Muth, Thilo

AU - Staege, Martin S.

AU - Daulatabad, Swapna V.

AU - Widera, Darius

AU - Zhang, Junpeng

AU - Meule, Adrian

AU - Honjo, Ken

AU - Pourret, Olivier

AU - Yin, Cong-Cong

AU - Zhang, Zhongheng

AU - Cascella, Marco

AU - Flegel, Willy A.

AU - Goodyear, Carl S.

AU - van Raaij, Mark J.

AU - Bukowy-Bieryllo, Zuzanna

AU - Campana, Luca G.

AU - Kurniawan, Nicholas A.

AU - Lalaouna, David

AU - Huttner, Felix J.

AU - Ammerman, Brooke A.

AU - Ehret, Felix

AU - Cobine, Paul A.

AU - Tan, Ene-Choo

AU - Han, Hyemin

AU - Xia, Wenfeng

AU - McCrum, Christopher

AU - Dings, Ruud P. M.

AU - Marinello, Francesco

AU - Nilsson, Henrik

AU - Nixon, Brett

AU - Voskarides, Konstantinos

AU - Yang, Long

AU - Costa, Vincent D.

AU - Bengtsson-Palme, Johan

AU - Bradshaw, William

AU - Grimm, Dominik G.

AU - Kumar, Nitin

AU - Bazanova, O. M.

AU - Москаленский, Александр Ефимович

N1 - Publisher Copyright: © The Author(s) 2019. Published by Oxford University Press.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Document recommendation systems for locating relevant literature have mostly relied on methods developed a decade ago. This is largely due to the lack of a large offline gold-standard benchmark of relevant documents that cover a variety of research fields such that newly developed literature search techniques can be compared, improved and translated into practice. To overcome this bottleneck, we have established the RElevant LIterature SearcH consortium consisting of more than 1500 scientists from 84 countries, who have collectively annotated the relevance of over 180 000 PubMed-listed articles with regard to their respective seed (input) article/s. The majority of annotations were contributed by highly experienced, original authors of the seed articles. The collected data cover 76% of all unique PubMed Medical Subject Headings descriptors. No systematic biases were observed across different experience levels, research fields or time spent on annotations. More importantly, annotations of the same document pairs contributed by different scientists were highly concordant. We further show that the three representative baseline methods used to generate recommended articles for evaluation (Okapi Best Matching 25, Term Frequency-Inverse Document Frequency and PubMed Related Articles) had similar overall performances. Additionally, we found that these methods each tend to produce distinct collections of recommended articles, suggesting that a hybrid method may be required to completely capture all relevant articles. The established database server located at https://relishdb.ict.griffith.edu.au is freely available for the downloading of annotation data and the blind testing of new methods. We expect that this benchmark will be useful for stimulating the development of new powerful techniques for title and title/abstract-based search engines for relevant articles in biomedical science.

AB - Document recommendation systems for locating relevant literature have mostly relied on methods developed a decade ago. This is largely due to the lack of a large offline gold-standard benchmark of relevant documents that cover a variety of research fields such that newly developed literature search techniques can be compared, improved and translated into practice. To overcome this bottleneck, we have established the RElevant LIterature SearcH consortium consisting of more than 1500 scientists from 84 countries, who have collectively annotated the relevance of over 180 000 PubMed-listed articles with regard to their respective seed (input) article/s. The majority of annotations were contributed by highly experienced, original authors of the seed articles. The collected data cover 76% of all unique PubMed Medical Subject Headings descriptors. No systematic biases were observed across different experience levels, research fields or time spent on annotations. More importantly, annotations of the same document pairs contributed by different scientists were highly concordant. We further show that the three representative baseline methods used to generate recommended articles for evaluation (Okapi Best Matching 25, Term Frequency-Inverse Document Frequency and PubMed Related Articles) had similar overall performances. Additionally, we found that these methods each tend to produce distinct collections of recommended articles, suggesting that a hybrid method may be required to completely capture all relevant articles. The established database server located at https://relishdb.ict.griffith.edu.au is freely available for the downloading of annotation data and the blind testing of new methods. We expect that this benchmark will be useful for stimulating the development of new powerful techniques for title and title/abstract-based search engines for relevant articles in biomedical science.

UR - http://www.scopus.com/inward/record.url?scp=85082592913&partnerID=8YFLogxK

U2 - 10.1093/database/baz085

DO - 10.1093/database/baz085

M3 - Article

C2 - 33326193

VL - 2019

SP - 1

EP - 67

JO - Database-The journal of biological databases and curation

JF - Database-The journal of biological databases and curation

SN - 1758-0463

M1 - 085

ER -