publicdata/nih-gov
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
nih-gov/www.ncbi.nlm.nih.gov/bionlp/Tools/simconcept

387 lines
No EOL
17 KiB
Text
Raw Permalink Blame History

<!DOCTYPE html>
<html lang="en" >
<head >
<meta charset="UTF-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<!-- Mobile properties -->
<meta name="HandheldFriendly" content="True">
<meta name="MobileOptimized" content="320">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<!-- Stylesheets -->
<link href="/research/bionlp/static/django_uswds/uswds/css/uswds.css" rel="stylesheet" />
<title>
SimConcept: A Hybrid Approach for Simplifying Composite Named Entities in Biomedical Text
</title>
<link rel="stylesheet" href="/research/bionlp/static/main/css/uswds.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/header.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/footer.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/form.css">
<!-- Labs template -->
<link rel="stylesheet" href="/research/bionlp/static/main/css/atoms.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/docsum.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/media.css">
<!-- Additional template -->
<link rel="stylesheet" href="/research/bionlp/static/main/css/journals.molecules.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/custom.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/journals.journal-page.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/iconic-glyphs.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/featherlight.min.css">
<link rel="stylesheet" href="/research/bionlp/static/main/css/styles.css">
<!--[if lt IE 9]>
<link rel="stylesheet" href="/research/bionlp/static/main/css/iconic-glyphs-legacy.css">
<![endif]-->
<!-- Some JS -->
<script src="/research/bionlp/static/main/js/jquery.js"></script>
<script src="/research/bionlp/static/main/js/modernizr.js"></script>
<script src="/research/bionlp/static/main/js/featherlight.min.js"></script>
<script src="/research/bionlp/static/main/js/custom.js"></script>
</head>
<body >
<div>
<a class="skipnav" href="#maincontent">
Skip to main page content
</a>
<header class="ncbi-page-header" role="banner">
<div class="prefix">
<span class="nih" title="National Institutes of Health">
<a href="https://www.nih.gov/" title="To NIH homepage">
<img src="/research/bionlp/static/base/images/nih-logo-header.svg" alt="NIH">
</a>
</span>
<span class="nlm">
<a href="https://www.nlm.nih.gov/" title="To NLM homepage">U.S. National Library of Medicine</a>
</span>
</div>
<div class="ncbi">
<!-- <abbr class="abbr">
<a href="https://www.ncbi.nlm.nih.gov/" title="To NCBI homepage">NCBI</a>
</abbr>
<span class="name">
<a href="https://www.ncbi.nlm.nih.gov/" accesskey="1" title="To NCBI homepage">National Center for Biotechnology Information</a>
</span> -->
<!-- <abbr class="abbr">
<a href="https://www.nlm.nih.gov/research/index.html" title="To DIR homepage">DIR</a>
</abbr> -->
<span class="name">
<a href="https://www.nlm.nih.gov/research/index.html" accesskey="1" title="To DIR homepage">Division of Intramural Research</a>
</span>
<div class="right">
<a id="in" href="/research/bionlp/accounts/login/?next=/research/bionlp/">Log in</a>
</div>
</div>
</header>
<!--app-specific header, something that might want to take full width of screen -->
<a class="skipnav" href="#maincontent">
Skip to main page content
</a>
<div class="breadcrumbs-container menu">
<div class="usa-grid-full">
<ul class="topnav" accesskey="4">
<li class="current">
<a href="/research/bionlp/" title="Home">
Home
</a>
</li>
<li class="separator"></li>
<li>
<a href="/research/bionlp/Zhiyong-Lu" title="Zhiyong Lu">
Zhiyong Lu
</a>
</li>
<li class="separator"></li>
<li>
<a href="/research/bionlp/News" title="Media">
Media
</a>
</li>
<li class="separator"></li>
<li>
<a href="/research/bionlp/Team" title="Team">
Team
</a>
</li>
<li class="separator"></li>
<li>
<a href="/research/bionlp/Research" title="Research">
Research
</a>
</li>
<li class="separator"></li>
<li>
<a href="/research/bionlp/Publications/" title="Publications">
Publications
</a>
</li>
<li class="separator"></li>
<li>
<a href="/research/bionlp/Tools/" title="Tools">
Tools
</a>
</li>
<li>
<a href="/research/bionlp/APIs/" title="Tools">
Web APIs
</a>
</li>
<li class="separator"></li>
<li>
<a href="/research/bionlp/Data/" title="Data">
AI Datasets
</a>
</li>
<li>
<a href="/research/bionlp/Visiting-us" title="Visiting us">
Visiting us
</a>
</li>
<li class="icon">
<a href="#">&#9776;</a>
</li>
</ul>
</div>
</div>
<!-- asign css class in case app will need to alter styles of this div -->
<div id="maincontent" class="usa-grid-full ncbi-base-page-container">
<div class="labs-pagecontent">
<div class="usa-width-one-whole">
<main class="usa-grid journals-lists">
<h3>SimConcept: A Hybrid Approach for Simplifying Composite Named Entities in Biomedical Text</h3>
<main class="usa-width-one-whole journal-container">
<div>
<div class="issue labs-docsums labs-content-box wrappall">
<h4>Authors: <a href="https://sites.google.com/site/chihhsuanwei/" target="_blank">Chih-Hsuan Wei</a>, <a
href="mailto:robert.leaman@nih.gov" target="_blank">Robert Leaman</a> and <a
href="/bionlp/" target="_blank">Zhiyong Lu</a> (PI)</h4>
<h4>Research highlights</h4>
<div class="usa-width-one-whole">
<p>
Here we propose a hybrid approach by integrating a machine learning model, named SimConcept, with a
pattern identification strategy to identify individual mentions from a composite named entity. More
specifically, we first trained and built a Conditional Random Fields model to detect the composite
mentions and subsequently identify the antecedent (e.g., colorectal) and conjuncts regions (e.g.,
adenomas and carcinoma) of a composite mention. Next, we manually developed four patterns to model
the six different types of composite mentions in our study. Finally, by applying our patterns to
those previously identified regions in the composite mention, individual mentions are generated in
our final output (e.g. colorectal adenomas and colorectal carcinoma).
</p>
</div>
</div>
<div class="issue labs-docsums labs-content-box wrappall">
<h4>Method overview</h4>
<div class="usa-width-one-whole">
<p>
SimConcept consists of two modules as shown in Figure 1. The first module consists of a conditional
random field model. In this module, the input mention is separated into tokens and each token
assigned labels according to the most likely sequence of states through the model. The second module
reassembles the tokens into individual mentions using a pattern identification method.
</p>
<div class="figure">
<img src="/research/bionlp/static/main/images/tools/SimConcept.png" width="350"/>
<span><b>Figure 1.</b> An overview of the SimConcept workflow.</span>
</div>
</div>
</div>
<div class="issue labs-docsums labs-content-box wrappall">
<h4>Results</h4>
<div class="usa-width-one-whole">
<p>
To evaluate our method, we used leave-one-out cross validation on the three sets (i.e., gene,
disease and chemical). Table 1 shows the results of our evaluation, where we see that the overall
performance is high for all three entity types. As mentioned in introduction, this study is aimed at
helping bioconcept normalization. We therefore applied SimConcept in GenNorm [21] and DNorm [18],
and evaluated on the test sets of BioCreative II gene normalization task [12] and NCBI disease
corpus [50], respectively (no normalized chemical corpus is available). To avoid training on the
test set, the training set for SimConcept excluded the test corpora for GenNorm and DNorm. As shown
in Table 4 and Table 5, using SimConcept can further improve the state-of-the-art performance for
1.17% in F-measure (P-value=0.02) for gene normalization and 1.34% in F-measure (P-value=0.03) for
disease normalization.
</p>
<table class="customtable">
<tbody>
<tr>
<td align="center"><strong>Bioconcepts</strong></td>
<td align="center"><strong>Precision</strong></td>
<td align="center"><strong>Recall</strong></td>
<td align="center"><strong>F-measure</strong></td>
</tr>
<tr>
<td align="center">Gene</td>
<td align="center">89.51%</td>
<td align="center">91.35%</td>
<td align="center">90.42%</td>
</tr>
<tr>
<td align="center">Disease</td>
<td align="center">87.92%</td>
<td align="center">85.07%</td>
<td align="center">86.47%</td>
</tr>
<tr>
<td align="center">Chemical</td>
<td align="center">87.44%</td>
<td align="center">84.71%</td>
<td align="center">86.05%</td>
</tr>
</tbody>
</table>
<span><b>Table 1.</b> The evaluation of SimConcept corpus.</span>
<table class="customtable">
<tbody>
<tr>
<td align="center"><strong>Tools</strong></td>
<td align="center"><strong>Precision</strong></td>
<td align="center"><strong>Recall</strong></td>
<td align="center"><strong>F-measure</strong></td>
</tr>
<tr>
<td align="center" class="best">GenNorm + SimConcept</td>
<td align="center" class="best">87.01%</td>
<td align="center" class="best">86.13%</td>
<td align="center" class="best">86.57%</td>
</tr>
<tr>
<td align="center">GenNorm</td>
<td align="center">86.72%</td>
<td align="center">84.09%</td>
<td align="center">85.38%</td>
</tr>
<tr>
<td align="center" class="best">DNorm + SimConcept</td>
<td align="center" class="best">80.91%</td>
<td align="center" class="best">79.23%</td>
<td align="center" class="best">80.06%</td>
</tr>
<tr>
<td align="center">DNorm</td>
<td align="center">80.69%</td>
<td align="center">76.85%</td>
<td align="center">78.72%</td>
</tr>
</tbody>
</table>
<span><b>Table 2.</b> The SimConcept contribution on gene, disease normalization performance.</span>
</div>
</div>
<div class="issue labs-docsums labs-content-box wrappall">
<h4>Downloads</h4>
<div class="usa-width-one-whole">
<p>
<a href="/CBBresearch/Lu/Demo/tmTools/download/SimConcept/SimConcept.zip"
target="_blank">SimConcept Source Code</a><br/>
</p>
</div>
</div>
<div class="issue labs-docsums labs-content-box wrappall">
<h4>Please cite</h4>
<div class="usa-width-one-whole">
<ul class="dot-list">
<li>Wei C-H, Leaman R, Lu Z. <a
href="http://delivery.acm.org/10.1145/2650000/2649420/p138-wei.pdf?ip=130.14.254.24&id=2649420&acc=OA&key=4D4702B0C3E38B35.4D4702B0C3E38B35.4D4702B0C3E38B35.9DBB59DEA6D22751&CFID=648750863&CFTOKEN=12918478&__acm__=1427473295_f84a1f644b896d02cdbff2edbb320044"
target="_blank">SimConcept: A Hybrid Approach for Simplifying Composite Named Entities in
Biomedicine</a>. Proceedings of the ACM Conference on Bioinformatics Computational Biology and
Health Informatics, Newport Beach, CA, 2014, p138-146
</li>
</ul>
</div>
</div>
</div>
</main>
</main>
</div>
</div>
</div>
<footer class="usa-footer usa-footer-big ncbi-footer" role="contentinfo">
<div class="usa-grid">
<div class="usa-row">
<div class="usa-width-one-half">
<div>
<div class="org-section">
<a href="https://www.hhs.gov/"><img class="usa-footer-logo-img hhs-logo"
src="/research/bionlp/static/base/images/dhhs-logo-white.svg"
alt="U.S. Department of Health & Human Services">
<span class="usa-sr-only">Department of Health and Human Services</span></a>
<a href="https://www.nih.gov/"><img class="usa-footer-logo-img nih-logo"
src="/research/bionlp/static/base/images/nih-logo-white.svg"
alt="National Institutes of Health">
<span class="usa-sr-only">National Institutes of Health</span></a>
<a href="https://www.nlm.nih.gov/"><img class="usa-footer-logo-img nlm-logo"
src="/research/bionlp/static/base/images/nlm-logo-letters-white.svg"
alt="National Library of Medicine">
<span class="usa-sr-only">National Library of Medicine</span></a>
<a href="https://www.usa.gov/"><img class="usa-footer-logo-img usagov-logo"
src="/research/bionlp/static/base/images/usagov-logo-white.svg"
alt="USA.gov"/>
<span class="usa-sr-only">USA.gov</span></a>
</div>
</div>
</div>
<div class="usa-width-one-half">
<div>
<p class="about-links">
<a href="https://www.nlm.nih.gov/research/index.html">About DIR</a>
<a href="https://www.nlm.nih.gov/web_policies.html">Web Policies</a></p>
</div>
</div>
</div>
</div>
</footer>
</div>
<!-- JavaScript -->
<script src="/research/bionlp/static/django_uswds/uswds/js/uswds.js"></script>
<script type="text/javascript" src="/research/bionlp/static/base/header.js"></script>
</body>
</html>
Reference in a new issue View git blame Copy permalink