You are on page 1of 3

145

RESOURCE REVIEW
DOI: dx.doi.org/10.5195/jmla.2018.280

Collaborators on the project in- are most interested in—references


Semantic Scholar. Allen Institute
clude CiteSeerX, a product of the and citations, methods as a limiter,
for Artificial Intelligence (AI2),
Colleges of Information Science and graphs and tables—without
2157 North Northlake Way, Suite
and Technology at Pennsylvania requiring a great deal of reading.
110, Seattle, WA 98103;
State University [4]. CiteSeerX, pre-
https://www.semanticscholar.org; With a growing need for re-
ceded by CiteSeer, attempts to in-
free. searchers and institutions to show
dex full-text portable document
impact, highly cited authors are
format files (PDFs) in computer
emphasized with influence scores,
and information science using
highly influential citations [5], total
completely autonomous citation
citations, a citations-per-year
indexing (ACI).
graph, and a citation velocity score.
INTRODUCTION Authors with fewer than 50 cita-
SEARCH RESULTS tions are merely noted as “<50”
Eagerly awaited by researchers for
years, concrete examples of artifi- Features of Semantic Scholar are in citations. Given sufficient citations,
cial intelligence–enabled search keeping with the mission to save author maps indicate those most
engines are beginning to emerge. the user time. By limiting results, influenced by an author and those
Founded by the nonprofit Allen Semantic Scholar renews a scien- with the greatest influence on an
Institute for Artificial Intelligence tist’s joy of discovery once again. author. The reference list brings
(AI2), Semantic Scholar began as a Searches that return tens of thou- deeper meaning to citations by
search engine for computer science, sands of results in Google Scholar showing where and how often a
geoscience, and neuroscience in and thousands in PubMed return a reference is cited in the paper
2015. In response to researchers’ few hundred in Semantic Scholar, through a display of the semantic
inability to keep pace with reading all directly relevant. Semantic context or contexts. Semantic
all of the publications in their dis- Scholar removes the long tail of Scholar attempts to combine con-
ciplines, the purpose of the project search results, allowing one to ventional citation metrics and alt-
is automated learning from text in quickly get up to speed on one’s metrics with the “cited by”
order to overcome information disciplines, while limiting the dis- function seen elsewhere in Web of
overload. traction caused by less relevant Science and Google Scholar, as well
research. Semantic Scholar also as links to tweets about citations.
This project is just one by AI2
saves time by its drive to provide
designed to fulfill the organiza-
full-text and mobile-enabled de- SCOPE
tion’s mission of “[artificial intelli-
sign. It truly is a search engine de-
gence] for the common good” [1]. Given the origins of the search en-
signed by researchers for
Semantic Scholar had ten million gine, neuroscience still predomi-
researchers.
articles prior to branching into bi- nates in the biomedical literature
omedical literature in 2017 [2, 3]. It that Semantic Scholar indexes. The
is reviewed here as an artificial in- CITATION ANALYTICS nature of searches may change as
telligence–based search engine now more citations are added.
The citation analytics features
poised to play a large role in health
graphically represent citation ve-
disciplines. Semantic Scholar is no-
locity and author influence scores INTEROPERABILITY
table among artificial intelligence
that help researchers pre-assess
search for the velocity with which Semantic Scholar does not offer an
quality, in much the same way that
it is being developed and the application programming interface
pre-assessed evidence levels aid
strength and knowledge of the de- (API), preventing interoperability
clinicians. Displays quickly visual-
velopment team. with other health care or biblio-
ize those elements that researchers

jmla.mlanet.org 106 (1) January 2018 Journal of the Medical Library Association
146 Review

DOI: dx.doi.org/10.5195/jmla.2018.280

metric systems. Users can copy and SIMILAR TOOLS ture impact, and at the time of this
paste citations in Bibtex, Endnote, Semantic Scholar aims to combat writing, it is not yet available to the
Modern Language Association the information overload and lack public [8]. Another artificial intelli-
(MLA), American Psychological of quality assessment that many gence–enabled search engine with
Association (APA), and Chicago researchers experience with Google a business focus, AlphaSense, has
Manual of Style styles, but citations Scholar’s keyword search. To ac- been available by paid subscription
cannot be exported. Users can cre- complish this, developers are cur- since 2010 [9].
ate reading lists if they sign in—via rently relying on preexisting
Facebook, Twitter, or Google—but indexing in PubMed and IEEE, CONCLUSION
those lists, also, cannot be exported. along with annotations provided
Authentication should be with the Despite limitations, Semantic
by medical subject experts. Articles
same account each time; that is, a Scholar achieves its aim of visually
without full-text links instead link
sign-in with Facebook will not link representing the important ele-
to a digital object identifier (DOI),
to a sign-in with Twitter. Users ments of a paper (key people, data
PubMed abstracts, or IEEE.
cannot create alerts to notify them analysis, and graphical representa-
of new search results. Like Google Scholar, Semantic tion of citations) without requiring
Scholar combs the web for citations a great deal of reading. As an in-
via an undefined algorithm, which troductory database on computer
SEARCH REFINEMENT
favors precision and full-text access science or neuroscience or one
Favoring simplicity of interface, over recall. Unlike Google Scholar, meant to rapidly display impact to
Semantic Scholar offers only a few it does not search behind paywalls. diverse stakeholders, it has value.
options for refining and sorting Thus, despite the developers’ stat- Transitioning to biomedical litera-
search results. It sorts only by rele- ed emphasis on quality, Sematic ture will take longer and require
vance and publication date. While Scholar’s inability to search li- greater oversight by subject ex-
it does allow truncation, it does not censed resources favors time, con- perts. While it works to overcome
support Boolean or phrase search- venience, and access. Hence barriers imposed by publisher
ing. Some limiters found here are researchers searching Semantic paywalls, Semantic Scholar’s great-
not found in other databases—such Scholar in its current form cannot est value may be its ability to visu-
as Data Set Used, Cell Type, and consider it a complete search of the alize open access research.
Brain Region—which reflects the background literature in their dis-
original audience targeted by the ciplines. REFERENCES
database.
PubMed now incorporates 1. Etzioni O. AI for the common good.
Other limiters provide few op- many of the visualization features MIT Technol Rev [Internet]. 23 May
tions. For example, Publication that Semantic Scholar explores in 2016 [cited 27 Jun 2017.
<http://events.technologyreview.com
Types limits currently only to jour- terms of citation graphs. While Se- /video/watch/oren-etzioni-ai-for-
nal article, review, study, meta- mantic Scholar has the added abil- common-good/>.
analysis, letters/commentary, clini- ity to limit to methods, it does not
2. Engleking C. Scientists are drowning,
cal trial, editorial, news, case re- offer a limiter comparable to Pub- artificial intelligence will save them.
port, and dataset, although these Med’s study type. Discover [Internet]. 11 Nov 2016 [cited
filter options are evolving with the 27 Jun 2017].
Other artificial intelligence– <http://blogs.discovermagazine.com/
addition of biomedical literature.
enabled search engines exist. Some d-brief/2016/11/11/semantic-scholar-
The Organism limiter refers to the artificial-intelligence/>.
have compared Semantic Scholar to
study population species only, not
the Memex project from NASA and 3. Nickelsburg M, Bishop T. Paul Allen’s
the infectious agents being studied.
DARPA [6] that searches the deep AI2 expands smart search engine Se-
Lists of extracted key phrases from mantic Scholar to neuroscience re-
web, though that project is not
citations often do little to further a search. GeekWire [Internet]. 11 Nov
available to the public. It is also 2016 [cited 27 Jun 2017].
search in the way that assigned
compared to Meta [7], now owned <https://www.geekwire.com/2016/p
Medical Subject Headings (MeSH)
by the Chan-Zuckerburg Initiative. aul-allens-ai2-expands-smart-search-
terms often can in PubMed. engine-semantic-scholar-neuroscience-
Meta was designed in 2010 with a
research/>.
greater emphasis on predicting fu

Journal of the Medical Library Association 106 (1) January 2018 jmla.mlanet.org
Review 147

DOI: dx.doi.org/10.5195/jmla.2018.280

4. College of Information Sciences and 6. Malone KL. Semantic Scholar. Techno- 8. Meta [Internet]. Meta [cited 21 Nov
Technology, Pennsylvania State Uni- philes Newscast 132 [Internet]. 30 Nov 2017]. <https://meta.com/>.
versity. CiteSeerX [Internet]. The Uni- 2015 [cited 27 Jun 2017].
versity [cited 21 Nov 2017]. <https://www.youtube.com/watch?v 9. AlphaSense [Internet]. Alphasense [cit-
<http://citeseer.ist.psu.edu/>. =95vndf_4t4w>. ed 21 Nov 2017]. <https://www.alpha-
sense.com/>.
5. Valenzuela M, Ha V, Etzioni O. Identi- 7. Sentence R. Chan-Zuckerberg Initiative
fying meaningful citations. In: Scholar- acquires AI-powered search engine Suzanne Fricke, DVM, MLIS, AHIP,
ly big data AI perspectives, challenges, Meta. Search Engine Watch [Internet]. suzanne.fricke@wsu.edu, orcid.org/0000-
and ideas: papers from the 2015 AAAI 25 Jan 2017 [cited 27 Jun 2017]. 0002-4412-9717, Animal Health Library,
workshop [Internet]. Association for <https://searchenginewatch.com/2017
Washington State University, Pullman,
the Advancement of Artificial Intelli- /01/25/chan-zuckerberg-initiative-
gence [cited 27 Jun 2017]. acquires-ai-powered-search-engine-
WA
<https://www.aaai.org/ocs/index.ph meta/>.
p/WS/AAAIW15/paper/download/1
0185/10244>.

Articles in this journal are licensed under a Creative


Commons Attribution 4.0 International License.

This journal is published by the University Library System


of the University of Pittsburgh as part of its D-Scribe
Digital Publishing Program and is cosponsored by the
University of Pittsburgh Press.

ISSN 1558-9439 (Online)

jmla.mlanet.org 106 (1) January 2018 Journal of the Medical Library Association

You might also like