Social Network Analysis and

Big Scholarly Data

Yogendra Singh
University Librarian
Swami Rama Himalayan University
Email - [email protected]
After this talk
• You should have a basic knowledge of Social Network Analysis

• You have an basic understanding of Big Scholarly Data

• You have an understanding of how Social Network Analysis can be applied to

Big Scholarly Data
Social Network
• Society is made of individuals (like wife and husband). They are known as
nodes/actors/vertices in Social Network parlance
• Individuals have relations
• Information flows between these relations
• These relations are known as ties/links/images in Social Networking
• Social network is a network (group) of individuals which have certain type of relations
for a particular information flow
Look at this picture
Now the big question is

• Who matters among this crowd? There could be different answers depending
upon different point of views

• The analysis of this type of social network using graph theory is called Social
Network Analysis

• Since Scholarly networks are the network of people (co-authors), it can well be
applied to large scholarly data or Big Scholarly Data
Big Scholarly Data (BSD)
•BSD refers to millions of scholarly records available today due to
tremendous changes in scholarly communication cycle

•BSD may include

• E-books, articles, reports, standards, patents etc., published by major commercial and not for
profit organizations - sciencedirect.com, tandfonline.com, doaj.org etc.
• Abstracting and Indexing databases- Scopus, Web of Science, EBSCO, Google Scholar
• Academic social networks- Academia, ResearchGate, Mendeley etc.
• Many other type of scholarly data
Three major scholarly data providers
Sl. No. Brand- Publisher Coverage No. of
name Records
1. Google Google Full Universe of Knowledge/ All 350+ million
Scholar Formats

2. Web of Clarivate Bibliographic Information including 90+ million

Science Analytics citations and other details including
abstract

3. Scopus Elsevier Bibliographic Information including 75+ million

Science citations and other details including
abstract
Big data analysis methods

•Statistical analysis
 Suitable for smaller datasets

•Scholarly text mining

 Can be used with big data

•Scholarly Network Analysis (or Social Network Analysis)

Scholarly Network Analysis/ Social
Network Analysis: Important measures
include Centralities

• Average path length

• Clustering coefficient

• Centralities
Average Path Length

• Average path length: Average distance of any two nodes in a

network is known as Average path length
Clustering Coefficient

• Average path length: Average distance of any two nodes in a

network is known as Average path length

• Clustering coefficient: is a measure of the degree to which nodes in

a graph tend to cluster together
Centrality Measures

• Average path length: Average distance of any two nodes in a network is

known as Average path length

• Clustering coefficient: is a measure of the degree to which nodes in a

graph tend to cluster together

• Centrality Measures: They measure how central (important) a node is in a

network
Network Centrality
 Which individuals (nodes) are important (Central)

 Measurement of importance is called Centrality in SNA

 Centrality may mean differently for different people and in different context
Why are Centrality and Centralization
Important?
• Access to information and ideas

• Interaction among members of the network

• Control the flow of information, resources, and other network content

• Visibility

• Ability to act together collectively

Multiple Ways to Calculate Centrality
• Degree

• Closeness

• Betweenness

• Eigenvector
Calculating Centrality
• Degree – Proportional to the number of other nodes to which a node is links –
Number of links divided by (n-1).
Calculating Centrality

• Degree – Proportional to the number of other nodes to which a node

is links – Number of links divided by (n-1).

• Closeness – The sum of geodesic distances (shortest paths) to all

other points in the graph. Divide by (n-1), then invert.
Calculating Centrality
• Degree – Proportional to the number of other nodes to which a node is
linked – Number of links divided by (n-1).

• Closeness – The sum of geodesic distances (shortest paths) to all other

points in the graph. Divide by (n-1), then invert.

• Betweenness – The extent to which a particular point lies ‘between’

other points in the graph; how many shortest paths (geodesics) is it on?
A measure of brokerage or gatekeeping.
Calculating Centrality
• Degree – Proportional to the number of other nodes to which a node is links – Number of
links divided by (n-1).

• Closeness – The sum of geodesic distances (shortest paths) to all other points in the graph.
Divide by (n-1), then invert.

• Betweenness – The extent to which a particular point lies ‘between’ other points in the graph;
how many shortest paths (geodesics) is it on? A measure of brokerage or gatekeeping.

• Eigenvector– A weighted measure of centrality that takes into account the centrality of other
nodes to which a node is connected. That is, being connect with other central nodes increases
centrality. E.g., secretary of powerful person. Google’s page rank algorithm is based on a
variation of this approach.
Network Analysis Tools Applied to BSD
Software/ Access Platform/ Language Description

CiteSpace/ Windows, IOS/ Visualizing and analyzing trends and patterns in scientific
Free Java
literature; knowledge domain visualization, best for WoS datasets

Gephi/ Windows/Linux/IOS Exploratory Data Analysis; Social Network Analysis; Link

Free Java
Analysis
iGraph/ Windows/IOS A collection of network analysis tools with the emphasis on
Free C/R/Python/Perl
efficiency, portability and ease of use

NetworkX / Windows/IOS Creation, manipulation, and investigation of the structures,

Free Python
dynamics,
and functions of complex networks
Pajek/ Windows/IOS Analysis and visualization of large networks having some
Free C/R
thousands or even millions of vertices
Types of Scholarly Networks could be
Generated by Applying SNA to BSD
• Co-Author Network
 Personal Network
 Organizational Network
 Geographic Network

• Co-Word Network
• Co-Citation Network
BSD Analysis Applications
•Scientific Impact Evaluation

 Article Impact

 Author Impact

 Journal Impact

 Institutional Impact
BSD Analysis Application - Academic
Recommendations

• Literature Recommendations

• Expert Recommendations

• Collaboration Recommendations

• Priority Recommendations
Scholarly Data Analysis: Steps
•Data Collection
 Download desired dataset from appropriate source

•Data cleaning
 Most difficult task as same name, institute, department is represented in different ways even by
the same individual

•Create graph using the data

•Use graph for further processing

Source and Software Needed
•An appropriate data source to download desired datasets
 I used Scopus to download research data of IIT Roorkee

•A software tool to clean data

 I used OpenRefine an open source software to clean the data, however, quite a bit was
done manually

•A software tool to create the graph

 An online tool Table2Net was used

•Process the graph for further obtaining necessary measures

 Gephi was used for this purpose
Steps in Analysing BSD through SNA
• Download a dataset

• Clean the data by some cleaning software such as OpenRefine and

Manually

• Create Graph File through some scientific network creating online tool
such as Table2Net or Scopus2Net

• Analyse that Graph file in Network Analysis software such as Gephi. You
can calculate all SNA measures using Gephi
Conclusion
• Application of Social Networking Tools to Big Scholarly Data is going to be big area of
interest to scientometricians as very large BSD is generated daily.
• These measures can be used to evaluate the authors, institutions, subject areas or
countries objectively.
• Special areas of interests, possible collaboration opportunities can be easily identified.
• As the impact of the publications can be easily identified, it will have great impact in
policy making.
• Librarians can also use SNA for analyzing in-house generated data such as circulation,
reference data, even footfall data.
THANK YOU