MeSH Now: Automatic MeSH Indexing at PubMed Scale

Yuqing Mao, Ph.D. and Zhiyong Lu, Ph.D. (PI)


MeSH Now is participating in BioASQ 2015 as baselines, see here for more info.

Research highlights

    MeSH indexing is the task of assigning relevant MeSH terms based on a manual reading of scholarly publications by human indexers. The task is highly important but also time-consuming and costly. It is reported that on average, it takes 2 to 3 months for a new article to be indexed upon entering PubMed [1]. We propose MeSH Now, a novel and scalable approach for MeSH indexing in real time. MeSH Now shows state-of-the-art performance in benchmarking evaluations, as well as top performance in BioASQ 2013/2014, a recently held global challenge on large-scale biomedical semantic indexing and question answering [2].

Method overview

    As shown in the figure blow, MeSH Now is an integrated approach that systematically merges inputs from multiple sources via its automatic learning framework [1,2].

flowchar

Results

    Below we show the performance of MeSH Now on three separate datasets and its comparisons to the MTI.

    Table 1. Evaluation of MeSH Now using micro-averaged precsion, recall and F-measure.

    Data set

    Methods

    Precision

    Recall

    F-measure

    NLM2007

    MTI 2014

    0.568

    0.525

    0.545

    MeSH Now

    0.610

    0.617

    0.614

    L1000

    MTI 2014

    0.561

    0.541

    0.551

    MeSH Now

    0.577

    0.626

    0.601

    BioASQ5000

    MTI 2014

    0.587

    0.559

    0.573

    MeSH Now

    0.614

    0.603

    0.608

Downloads

Publications


Last Update: Jan 22, 2015

Quick Links:
tmTools
PubTator
BioASQ

 

Quick Download:
Paper (PDF)
Citation (RIS)
RESTful API Download - coming soon

 

Contact Information:
Dr. Zhiyong Lu
MSC3825 NCBI/NLM/NIH
Bldg 38A, Rm 1003A
8600 Rockville Pike
Bethesda, MD, 20894
Tel: 301-594-7089
Fax: 301-480-2288
zhiyong.lu@nih.gov