Invited talk on "why altmetrics?" at the information day "Bibliometrics, Scientometrics & Alternative metrics: which tools for which strategies?”, Association des directeurs et personnels de direction des bibliothèques universitaires et de la documentation (ADBU), 1st April 2015, BULAC, France (Paris)
2. Outline
• The
importance
of
altmetrics
• Introduc4on
of
altmetrics
(concept,
tools
and
data
sources)
• What
do
we
know?
• What
does
it
show?
• What
we
can
do?
• Problems
&
Opportuni4es
1
3. Why
altmetrics
is
important?
• Different
ways
of
measuring
impact
of
research:
pros
&
cons
•
Tradi4onal
metrics:
peer
review
and
cita4on
analysis
• Novel
web-‐based
metrics:
‘Altmetrics’
or
‘social
media
metrics’
Limitations:
• Scope: partial vs complete impact/quality of research
• Format: some limited vs diverse researh products (articles and
reviews vs dataset, blog, software, etc.)
• Speed: lots of time to accumulate vs the real-time impact
• Audience: measuring only the impact of research by researchers
(known as scientific impact) vs broad audience such as use of
research by general public (societal impact?)
• Real time impact: superficial use?
• Reliability: gaming, manipulation or boosting the impact?
• Producers: human or robot?
2
4. Then what is the best approach for measuring
impact/quality of research?
still unanswered question but there is no single way!
3
5. What
is
Altmetrics?
• There
is
no
exact
defini4on
of
altmetrics
(from
the
conceptual
point
of
view)
• “It
is
a
good
idea
but
a
bad
name”
(Rousseau
&
Ye,
2014)
•
Diversity
of
terms
suggested:
‘social
media
metrics’,
‘influmetrics’,
etc.
• Altmetrics
are
seen
as
metrics
about
ar4cles
(ar4cle-‐level
metrics)
vs
journal
impact
factors
(journal
level
metrics)
•
first
introduced
in
the
‘Altmetrics
manifesto’
By
Preim
et
al.
In
2010
• Some
consider
all
the
views,
downloads,
readerships,
men4ons
in
social
media
and
news
media,
etc.
as
altmetrics
•
Refers
to
different
ac4vi4es
in
different
plaZorms:
blogs,
Twi[er,
Facebook,
Wikipedia,
reference
management
tools,
etc.
4
6. Altmetrics definition
Altmetrics
refers
to
the
men4ons
of
scien4fic
outputs
in
‘social
media’
(e.g.
Twi[er,
Facebook,
blogs,
etc.)
or
crowdsourced
tools
(e.g.
Mendeley)
or
any
online
ac4vity
around
research
products
captured
by
altmetrics
tools
5
7. What
are
the
tools
and
data
sources
for
‘altmetrics’
Diversity
of
tools
are
available
among
them
are:
• Impact
story
• Altmetric.com
• PloS
one
• Plum
Analy4cs
• Mendeley
some
offers
open
API!
single
vs
divers
metrics
are
avaialable
by
them
6
9. • A free reference management tool
• Database of scientific outputs
• More than 2.8 million users
• Usage statisitics based on users
• Open API
What type of altmetrics data?
10. Distribution of readerships by users
across LR fileds
9
• What are the most common types of users in Mendeley?
11. Limita.ons
of
Altmetrics
tools
• Differant
metrics
available:
readerships,
tweets,
FB
shares,
comments,
blogging,
etc.
• Limita4ons:
• No
clear
meaning
of
these
metrics
• Manipulability
• Difficult
scalability
and
data
collec4on
(although
API’s
are
available)
• No
normaliza4on
of
indicators
• Low
level
of
data
standardiza4on
10
12. • Main research topics, so far
– Coverage
– Correlations
– Content analysis
– Data problems, quality & validity
11
What do we know about altmetrics?
13. Coverage of publications
Data Source
Papers
with
metrics
%
papers
without
metrics %
Mendeley
readers
12,362
62.6
7,392
37.3
Twitter
324
1.6
19,448
98.3
Wikipedia
mentions
289
1.4
19,483
98.6
Delicious
bookmarks
72
0.3
19,700
99.7
12
• Which altmetrics data source & tools have the highest
coverage of WoS publications?
a random sample (20,000) of WoS publications from all fileds
of science (2005-2011) using Impact Story:
Mendeley has the highest coverage (62.6%)
14. - OK, for Mendeley (>70%)
- Low for
other sources
- Increasing
over time
13
Coverage of publications?
(Robinson-García et al, 2014)
(Costas et al, 2014)
15. (Cor)relations with citations
• Moderate correlations with citations for Mendeley
• Weak for the other sources (F1000, Twitter, blogs,
news, etc.)
• Also not very good at filtering highly cited
publications except for Mendeley
14
Zahedi, Costas & Wouters, 2014;
Waltman & Costas, 2014
General Precision-recall curves for JCS (blue line) and
total altmetrics (green line) for identifying PPtop10%
most highly cited publications
16. • There are some remarkable patterns (Costas et al, 2015):
– Twitter, stronger in Social Sciences and General medicine, lower
in Natural Sciences and Humanities
15
Coverage by fields
17. Content
analysis
• Analysis
of
the
4tles
and
abstracts
of
publica4ons
with
altmetrics
[Costas
et
al,
2014;
Zahedi
&
Ness,
2014]
• Distribu4on
of
Cita4ons
and
Altmetrics
by
• Disciplines
(Subject
Categories)
• Topics
(terms
in
the
4tles
and
abstracts)
• VOS
viewer
(www.vosviewer.com)
16
18. What are the fields with more density
of readerships vs citations?
Readership activity vs.
citation activity:
- Social Sciences
- Humanities
26. - Inconsistencies across different altmetrics providers
(due to their different data collection process, time,
etc.)(Zahedi, Fenner & Costas, 2014)
- fluctuation in Mendeley coverage and readership
counts over time and through different retrieval
strategies (Bar-Ilan, 2014; Zahedi, Haustein,
Bowman, 2014)
- Duplicates, different versions of the same document,
data quality (in complete or incorrect
metadata), retrieval errors (DOI and title-author
search?), user profile update?
25
Data problems summary
27. What are the opportunities?
– as an indicator of hot/popular topic discussed
– as an informative tool of identifying usage pattern by different
user category (student vs researchers vs other professions or
general publics): informing other types of impact (Societal,
cultural vs scientific impact)
– as an indicator for evaluating scientific impact: may be for
Mendeley but not yet for Twitter, blogs, news media, etc.
– as a complementary tool in measuring research impact
especially for some fileds with low coverage in citation databases
But still not known yet: whether altmetrics shows research impact?
visibility? Attention? Buzz, popularity or noise?
But many unanswered questions &
Still more research needed!
26