Generating Insight Using Text Network Visualization

By Dmitry Paranyushkin

Elevator Pitch

In this session I will demonstrate how we can use network graphs to generate insight, new ideas, and eureka moments. The approach is based on identifying the structural gaps in the graph and bridging those gaps by introducing the new connections.

Description

A gap is an opportunity. In this talk we will demonstrate how we can use graph theory and network visualization to detect structural gaps in knowledge graphs and text networks in order to generate insight and new ideas. For example, knowing a gap between what people search for and what they actually find on Google can be a very powerful use case for SEO and marketing applications. Knowing a gap in your research area may reveal a potentially interesting innovation.

Our method is based on the text network analysis and visualization algorithm which represents any text as a network and identifies the most influential words in a discourse based on the terms’ co-occurrence. Graph community detection algorithm is then applied in order to identify the topical clusters, which represent the main topics in the text as well as the relations between them — providing a visual insight about the main topics, on par with other topic modeling methods like LDA.

Graph topology is then analyzed in order to find the structural gaps between the different topical clusters. These gaps are then used to identify the parts of the discourse where the connections are lacking, therefore highlighting the areas where there’s a potential for new ideas and innovative research. This same approach and methodology can also be used for other types of graphs as well.

The demonstration will be made using the open-source text network analysis tool InfraNodus which uses this methodology. We will also present the Neo4J data structure model used by InfraNodus to store and retrieve data and talk about the challenges that arise when storing large text networks created in multiple contexts by multiple users.

Notes

The method has been developed over the last 7 years and has been implemented into InfraNodus software since 2015. We have numerous use cases published online on our website — https://noduslabs.com — where we have successfully demonstrated how this approach can be used in the context of research, marketing, education and art. Using Neo4J as the graph database underpinning our technology we’ve encountered several technical issues with scalability and have successfully resolved them. So this talk will be interesting both for those who want to learn about the graph method as well as the technicalities of its implementation.

If the descriptions of GraphConnect talks are available online in time for our presentation we will use them as source material to demonstrate how our approach can be used to identify the main topical clusters and structural gaps in the content of the conference itself.