=Paper=
{{Paper
|id=None
|storemode=property
|title=ClinicalKey: Terminology driven Semantic Search
|pdfUrl=https://ceur-ws.org/Vol-897/demo_9.pdf
|volume=Vol-897
|dblpUrl=https://dblp.org/rec/conf/icbo/ArabandiM12
}}
==ClinicalKey: Terminology driven Semantic Search==
https://ceur-ws.org/Vol-897/demo_9.pdf
ClinicalKey: Terminology driven Semantic Search
Sivaram Arabandi1*, Helen Moran1
1
Smart
Content
Strategy,
Elsevier
Health
Sciences,
USA
2.2 Smart Content
1 INTRODUCTION
Smart Content is content with a high level of structure, cre-
In this Information Age, we have a variety of sources to
ated by annotating text with a standardized terminology.
fulfill our information needs. More and more of the data is
The terminology, with its logical structure, adds the seman-
available online, and search engines such as Google, pro-
tic meaning – what the content is about and how the differ-
vide us with powerful tools to find specific pieces of infor-
ent pieces of content relate to each other.
mation. However, the data is growing exponentially result-
ing in an Information Overload [Bergamaschi & Guerra].
Rindflesch and Aronson describe the use of Natural Lan-
This phenomenon is all too common in the healthcare do-
guage Processing (NLP) with UMLS to extract usable se-
main too where clinicians are spending increasing amounts
mantic information from Medline abstracts. In creating
of time filtering out useless information to find what they
Smart Content, natural language processing is used by a
are looking for.
Query Parsing Engine (QPE) to identify concepts in Else-
vier’s medical corpus consisting of more than 400 top jour-
ClinicalKey is an innovative new online resource built on
nals, over 700 books and multimedia, as well as expert
Elsevier’s Smart Content – searchable journal, book, image
commentary, MEDLINE abstracts and select third-party
and video content tagged to EMMeT (Elsevier Merged Med-
journals. Once the QPE has identified the term labels – syn-
ical Taxonomy). It is designed from the ground up to pro-
onyms, acronyms, abbreviations, etc. in EMMeT – mapping
vide improved access to clinical information, providing
rules are applied by a Concept Mapper (CM) to create a
comprehensive, trusted clinical answers quickly (Figure 1).
searchable index of concepts with relevancy scores and
links to the originating documents. The resulting concept
index is also used to generate RDF satellites that populate
2 SMART CONTENT PLATFORM
Elsevier’s Linked Data Repository (LDR).
2.1 EMMeT
EMMeT is a clinical terminology model that is being devel- 2.3 ClinicalKey
oped to serve as an authoritative reference for clinical terms
The first product to utilize Elsevier Smart Content, Clini-
and the relations between them. EMMeT is envisioned as a
calKey is a web-based application for clinicians in the hos-
multi-product, re-usable ontology resource i.e. it will serve
pital setting. It provides a simple search interface for users
the needs of multiple applications. It is based on UMLS, and
to enter text that is then processed by the QPE. The QPE
as of March 2012, has over 1 million concepts and 3 million
interprets the search text and suggests EMMeT terms for
synonyms. The concepts originate from a subset of UMLS
auto-complete [Figure 2]. Along with the term labels, addi-
terminologies mainly SNOMED CT, RxNORM, ICD-9, and
tional information such as the Semantic Type of the term is
CPT; from Gold Standard drug database; and a number of
also displayed for disambiguating similar sounding terms.
terms have been introduced locally for aiding search.
Furthermore, the term-level relations from EMMeT are used
Figure 1: ClinicalKey based on Smart Content
1
Arabandi et al.
Figure 2: Search text
with Auto-complete
to suggest related searches that might be of interest to the
user. A number of additional tools such as the ability to save fre-
quent searches, search results, presentation maker, etc. are
The search process uses the indexed content and relevancy also available to users. ClinicalKey launched in April 2012
scores to retrieve the relevant answers. The results are dis- and is available at:
played to the user as a ranked list of titles – journal articles, https://www.clinicalkey.com/
book chapters, etc. [Figure 3]. The results are also catego-
rized into additional facets such as journals, books, images, REFERENCES
videos, etc. and into clinical domain categories (with sum- Bergamaschi S., Guerra F. (2010). Guest Editors' Introduction: Infor-
mary metrics), which can be used for further filtering of mation Overload. Internet Computing, IEEE.
results. The preview functionality allows users to quickly Rindflesch, T. C. and Aronson, A.R. (2002). Semantic Processing for En-
review the most highly ranked suggested paragraphs from a hanced Access to Biomedical Knowledge, in Real World Semantic
given article or book chapter for relevancy before accessing Web Applications, IOS Press, 157-72.
the full text.
Browse
&
Search
Facets
Tools
Results
with
Preview
Figure 3: Results with Facets, Preview and Tools
2