VISHLESHAKEE is a Social Media Analysis Tool developed at Open Source Intelligence Lab, Indian Institute of Technology, Guwahati, India. The tool is developed using open source technologies and uses publicly available data for analysis. The current version 2.0.0 of the tool considers open-sourced Twitter (a social micro-blogging platform) data collected via its provided API as the data source.

VISHLESHAKEE acts as a generalized information platform by providing the general public, knowledge about recent topics of discussion on social media. The tool denotes the topics in terms of Twitter hashtags and categorizes them into classes of Trending, Security, and Communal. Also, the associated public sentiments and locations of discussion can be easily visualized from the tool. Besides acting as a generalized information platform VISHLESHAKEE is also empowered with specialized modules for in-depth analysis of various social media contents. Access to such in-depth analysis adds values to the knowledge base of different law enforcement, homeland security, and other relevant agencies about the monitored topic of discussions. The latest version of the tool has 5 such modules, namely:-

• Location Monitor
• Trend Analysis
• Historical Analysis
• Network Analysis
Network Analysis Module

The Network Analysis module of Social Media Analysis Tool VISHLESHAKEE provides the functionality of analyzing different types of networks obtained from the system as well as networks that are being imported from the external source. As, the social media data is highly relational analysing them from the relational perspective can help gaining a better insight. The networks currently being supported by the system can be seen from the Network choice selector in the query panel. The module can generate a network given by the user with the required mandatory information and can perform various analysis on them. The rest of the manual provide the steps to be followed for generating a network and doing suitable analysis on them.

Minimum User Device Specification: 4 GB (RAM)

Recommended User Device Specification: >= 8 GB (RAM)

Network Generation

Network Generation from User Inputs

For generating the network, a user is required to provide the following mandatory information.



1. Query: The query is the input given by the user it may be a hashtag or a mention or a user ID, or a keyword. Sample Inputs: #JNU, @ndtv, narendra modi, cricket
2. From Date: Starting date of the range query
3. To Date: Ending date of the range query
4. Network Size: The upper bound on the number of nodes to be generated for the query. Say, 50 or 100.
5. Network Choice: The selector provides the facility to generate different networks. One need to select appropriate network to be generated for the given query

It is to be noted that the choice of the network should be compatible with the given query. i.e. if the given input query is a hashtag then the following networks should be only chosen for the network generation.

1. Hashtag-Hashtags.
2. Hashtag-Mentions.
3. Hashtag-User

Similarly, if the given input is a mention then the following networks are valid:

1. Mention-Mentions
2. Mention-Hashtags.
3. Mention-Users.

Once, all the inputs are being provided the user is required to click on the Generate Network button placed by the side of the input box in the query panel. This will result in the graph to be generated. A card will show up down the query panel with the basic information about the network being generated. The loader inside the card appears marking the process running for generating the graph. Wait, till the loader disappears from the card.

Network Generation from External File Upload

A network can also be generated by importing data from external source. The module currently accepts .csv files as edgelist (in the format of [src,dst,count]) as an external source for network generation.



Once, the loader disappears, The user is required to click on the view-graph text present on the card. Clicking on the text will show up the generated graph to the user in the Graph View Panel. Let the graph appear completely in the panel with all the edges and nodes rendered and wait till the graph becomes stable. It will take few seconds for the graph to be stable depending upon your system specifications. Note: It is to be noted that performing some analytical operations on the unstable graph may result in failing operations.)

Once, the graph is being generated a number of operations can be performed upon that. The analytical operations supported by our module with the steps to be followed are given below: -

For, going for performing the analytical operations on the graph the user needs to click on the PERFORM ANALYSIS button present in the right-hand side of the screen. Clicking, on the button will provide the user the choice buttons for the analytical operations:

Centrality

In social networks, centrality is an indicator of identifying important nodes in a network. Our module provides the following variants of centrality:




1. Degree Centrality
2. Eigen Value Centrality
3. Betweeness Centrality
4. Page Rank Centrality


For performing the centrality operation on the network. The following steps need to be followed:

1. Click on the blue colored Centrality Tab
2. Select the algorithm choice for the centrality finding provided via the radio buttons.
3. Click on the EXEC button.


Upon success of the query the resultant graph will be displayed in the same panel. Zoom out the division for letting the graph to be displayed as a whole.

Community Detection

In social networks Community Detection aims to group similar nodes together, thus discovering hidden communities present. Our module provides the following variants of Community Detection




1. Async Fluidic
2. Label Propagation
3. Grivan Newman


Steps to be followed for finding out the communities in the graph

1. Click on the community Detection Tab
2. Select the appropriate algorithm choice, if you select Async Fluidic as the algorithm you need to provide the number of communities to be generated
3. Click on the EXEC button.


In the above Community Detection sample output each color defines a community where the nodes with similar color belongs to the same community. Upon successful query a graph will be displayed in the same panel with communities in different colors.

Shortest Paths

Shortest path aims to find a shortest path between a pair of nodes in the network. Our module provides the following variants of Shortest Path.

1. Shortest Path
2. K-Shortest Path
3. All Possible Shortest Path per depth


Steps for finding the shortest path in the graph is as follows:

1. Click on the Shortest Path Tab
2. Provide Source and Destination node. One can provide this information directly in the query box or by clicking on the nodes in the graph. The selection of two nodes from the graph: First node can be selected by a direct single click while the selection of the next node requires Cmd+click on the node.
3. Select the Algorithm choice and Click on the EXEC button


On successful query the source and target node will be enlarged also the edge between them will be bold. If any intermediate node exists between the two nodes, then the intermediate node will also be enlarged colored in different color from the source and destination node.

Link Prediction

Link prediction aims to predict future connections between nodes based upon their present neighbbourhood. Our module provides the following variants of link prediction.



1. Adamic Adar
2. Jaccard Coefficient

The steps to be followed for the link prediction is as follows:

1. Click on the link prediction Tab
2. Select the Algorithm choice.
3. Select a node from the graph.
4. Provide the number of top k links to be predicted.
5. Click on the EXEC button

Upon successful query the query node and the predicted nodes will be highlighted in different color and the predicted edges will be appearing in dark black color.

Union of Networks

Union aims to combine different networks into a single network. Prerequisite: Generate two graphs from the Network Generation Panel. Make sure both the graphs are selected.



1. Select two graphs
2. Click on Union View Panel.
3. Click on EXEC Button

Upon successful query a graph with the union of two graphs will be displayed. Here in the Figure: Union Network node with similar color belongs to one of the input networks and nodes with the red color describes the nodes that are common to all the input networks.

Intersection of Networks

Intersection aims to find out the common nodes in a given number of input networks. Perquisite is as same as the union operations. Steps:



1. Select two graphs
2. Click on Intersection View Panel.
3. Click on EXEC Button

Upon successful query a graph with the intersection of two graphs will be displayed. Here in the Figure:Intersection Network nodes with red color determines the common nodes where as nodes with violet color denotes the nodes that are connected with the intersecting nodes and the nodes in dim denotes that are not satisfying the earlier criteria.

Difference of Networks

Difference of networks aims to discover uncommon nodes from a network with an another network. Upon successful query a graph with the intersection of two graphs will be displayed. Perquisite is as same as the union operations. Steps:



1. Select two graphs
2. Click on Difference View Panel.
3. Input the network IDs as per the ordering in which you want to perform the difference.Say, if you want to perform difference of network 1 from network 2, then you should add network 2 first and then network 2
4. Click on Perform Difference Button.

Deletion of a node from the Network

A node can be deleted from the network by double clicking on the node. It will result in the deletion of the node alongwith its related edges. The deletion can be made by double clicking on the node.



Distributed Processing of Large Scale Networks with Apache Spark

VISHLESHAKEE faciliates processing of large scale network with Apache Spark distributed computing framework.



;
• User Analysis

VISHLESHAKEE is in its growing phases. With prospects of further addition of much detailed and fine-grained analytical components as a result of research and development activities from the lab, VISHLESHAKEE will emerge as an all-round social media analysis tool.