View
62
Download
0
Category
Preview:
Citation preview
Research objectives:
To study:
• Mendeley’s coverage of JASIST articles over time
• Readership counts over time
• Why JASIST?
2
Data collection
• All JASIST articles published between 2001 and 2011 and indexed by Web of Science were retrieved
• Info on a few missing articles was filled in manually
• Missing DOIs were supplied
• Citation data from WOS, Scopus and GS were collected twice, once in 2012 and once in 2014
3
Mendeley searches
• In April 2012 and August 2013 all the searches were conducted manually through the website
• Title searches
• Special characters excluded
• ? “ : & …
• Search results matched against full title and author
• Seemingly wrong publication source was double checked (DOI or abstract)
• Readership counts of multiple records for a given article were combined
4
Mendeley searches (2)
• In 2014 we used Mike Thelwall’s Webometric Analyst (http://lexiurl.wlv.ac.uk/) for title searches
• Carefully compared with manual searches
6
JASIST 2001-2011 • 1645 articles
7
# articles on Mendeley % of total (1645) Total readers
Apr-12 1,600 97.3% 16,436
Aug-13 1,540 93.6% 24,851
Apr-14 1,453 88.3% 32,968
May-14 1,607 97.7% 39,635
05,000
10,00015,00020,00025,00030,00035,00040,00045,000
Ap
r-1
2M
ay-1
2Ju
n-1
2Ju
l-1
2A
ug-
12
Sep
-12
Oct
-12
No
v-1
2D
ec-1
2Ja
n-1
3Fe
b-1
3M
ar-1
3A
pr-
13
May
-13
Jun
-13
Jul-
13
Au
g-1
3Se
p-1
3O
ct-1
3N
ov-
13
Dec
-13
Jan
-14
Feb
-14
Mar
-14
Ap
r-1
4M
ay-1
4
Total readers
1,350
1,400
1,450
1,500
1,550
1,600
1,650A
pr-
12
May
-12
Jun
-12
Jul-
12
Au
g-1
2Se
p-1
2O
ct-1
2N
ov-
12
Dec
-12
Jan
-13
Feb
-13
Mar
-13
Ap
r-1
3M
ay-1
3Ju
n-1
3Ju
l-1
3A
ug-
13
Sep
-13
Oct
-13
No
v-1
3D
ec-1
3Ja
n-1
4Fe
b-1
4M
ar-1
4A
pr-
14
May
-14
# articles on Mendeley
Readership counts – closer look
• For 1102 out of the 1645 articles (67%) readership counts increased monotonously
• Only 6 articles were never located
• All six were cited more than once
• 543 non-monotonous cases
• Maximum decrease in readership count: 240 • “The link-prediction problem for social networks”
• Identical title in conference proceedings and in JASIST
• Wrong source attribution is sometimes the source of large changes in readership counts
• Other case: found through title search in April 2014, not retrieved in May 2014, but found through URL saved in April – retrieval problems?
• All items not located in April or May 2014 were double and triple checked
9
April 2012 vs. May 2014 • April 2012: 1600 out of 1645 articles
• May 2014: 1607 out of 1645 articles
• Are the same articles missing from both sets?
• Only 6 identical articles!
• The six that were never retrieved
• The 45 articles not found in April 2012:
• 320 readers in May 2014
• Most read article is from 2010 with 63 readers in May 2014
• The 38 articles not found in May 2014
• 325 readers in April 2012
• Most read article is from 2007 with 44 readers in April 2012 and 67 readers in August 2013
10
What happens with readership counts over time?
• Users who bookmark the specific item and include it in their Mendeley library
• What happens when the user deletes her account?
• What happens when the user deletes an item from her library? • And if she later bookmarks the same item again???
• How do these actions influence the readership counts?
• The clustering process run from time to time by Mendeley also affects readership counts
• Attribution of publication source
11
Two approaches
1. Once a reader always a reader
– This should result in non-decreasing readership counts
2. Only current readers are readers
– This can explain fluctuations in the counts
• A mixture of the two is also possible
– Deleted account deletes all records
– Deleted items in an existing account do not decrease reader counts
12
Suggested solution • Provide the number of readers of the item per month
• Total number of unique readers who bookmarked the item
• This would greatly increase Mendeley’s value as an altmetric
• Easier to report the number of Mendeley readers with a timestamp
• Allow to study interest trends, assuming that users delete items they do not find interesting
• In any case a transparent and consistent solution is needed
13
Recommended