Sometimes we want to get an overview of a new subject, discourse or a collection of materials (e.g. customer feedback, social media posts or news) in order to know where to start learning about it or simply to better understand what it's about.
InfraNodus is great for this job as it allows you to look into a certain subject through the prism of a network, to see the most important parts that comprise the whole and how they are interrelated.
Step by Step Guide:
Stage 1: Think of a Topic (Zoom In, Focus)
1. Think of a general subject you're interested in. For example, "Neurodynamics" or "Coronavirus" or "investment strategies" or "spiritual awakening" or "Simpsons" or "BMW".
2. Think of what would you like to find out about it. For example, what exists out there in general on this subject, first? What's the discourse around this topic?
3. Open InfraNodus.Com Google app
4. and type in the search query into the system. For instance, "text network analysis".
Stage 2: Get an Overview (Zoom Out, Explore)
5. The system will analyze and visualize the first 40 Google search results for this query as a graph:
You will notice that the search query items are removed from the graph (top right), so they don't overload the search results.
The graph itself shows you the most influential node-words based on their betweenness centrality (BC) measure (can be changed in the Settings). The bigger the node, the higher is the BC, the more influential it is.
The node / words that are closer related together (or co-occur) in the same search results are closer to each other on the graph and form distinct communities (indicated with a color) — these are the topics.
6. If you click on the Analytics panel (bottom left) > Essence Tab you will see the most important information presented to you: the most influential words and topics in this discourse:
7. In our case, the most influential words, apart from "text", "network" and "analysis" which were removed are
method, visualization, mining, visualize, topic
So we can see that when people talk about text network analysis it's closely related to text mining and to visualization.
Actionable insight: If we want to learn more about this topic we should look into text mining, topic mining, and network visualization techniques
8. We will also see that in this discourse people tend to talk about the following topics:
1: post part visualize
2: visualization paranyushkin tool
3: topic identify structural
4: plant bias base
They are indicated with the same colors and those nodes are closer to each other on the graph.
Stage 3: Zoom In, Explore
9. If we click on each of those topics we will be able to see which context they are being used in. For instance, clicking "plant bias base" shows us the part of the text that has the highest concentration of these terms:
10. Once we explore a few topics this way, we will get a general understanding of what this discourse is about.
Stage 4: Find the Structural Gaps, Explore the Nuance
(Zoom Out, Focus)
11. Use the Analytics > Gap Insight to reveal the structural gap: topics that should be connected but are not. This may indicate some of the ideas present in the existing discourse that could be developed further. It might also bring you to Google search results that may be interesting for you.
12. Next, we might be interested in what's hiding behind the obvious concepts that we already know. So then we can select the most influential nodes and remove them from the graph. We can also use the "Reveal Non-Obvious" button under the most influential nodes to see what's hiding behind them.
Try it yourself with this public demo graph:
Please sign in to leave a comment.