In this brief tutorial, we introduce the basic workflow you can use to get an overview and insight into your data using text network visualization generated with InfraNodus and built-in AI tools.
After you read the tutorial, you'll be able to apply the same approach to your own data to not only get an overview of discourse but also to generate new interesting ideas. You can use your own text or any existing discourse (e.g. Google search results or AI-generated facts).
Text Graph Workflow Tutorial:
The conceptual framework for research implemented in InfraNodus is based on inducing cognitive variability and ecological thinking. We do this by encouraging you to alternate between scales as you study a discourse (i.e. zooming in / zooming out) as well as by helping you shift your attention between focus and exploration. This workflow below exemplifies this approach.
Step 0: Create a New Graph and Choose the Editor Mode.
There are 3 options available:
1. Analyze a Text — upload or copy and paste an existing text, detect the main topics and the relations between them. This is the option we demonstrate in the workflow below.
2. Develop an Idea — start with an idea, a research question, or an abstract. Then use the AI assistant to explore it further. If you decide to choose this option, please, go to 🎥 Live AI Ideation Workflow: Develop Ideas using GPT-3 and Text Visualization
3. Explore a Topic — start with a keyword or a topic, and either import the Google search results to understand the context around or use the live AI ideator to generate interesting facts, then add them to the graph. If you want to try this option, go to 🎥 Explore a Topic using GPT-3 AI and Text Data Network Visualization
Below we will demonstrate how you can proceed if you choose the first option: Analyze a Text. You can also use this approach below with the other two options once you populate your graphs with content, in order to analyze the discourse generated. For instance, an easy way to populate your graph with content is to use the Google import app.
Step 1: Add a Text
Copy and paste a text into the editor or import an existing text to visualize it as a graph: a PDF or a text document, Google search results, or Twitter conversations around a certain hashtag.
If you import Google search / Google scholar results, you can simply make a search query that relates to your interest and visualize all the results.
Step 2: Generate a visual overview
Task: Zoom Out | Explore
The network visualization of text will show you the main ideas in your discourse, the main topics, and how they are related (here's a link to the paper that describes our methodology). We found that visualizing a text as a graph helps us think more in terms of connections (learn more in our research).
The bigger nodes on the graph have higher influence in the discourse network (based on betweenness centrality).
The nodes that have the same color belong to the same topical cluster — they are also positioned closer to each other on the graph (we use force-atlas algorithm). If the nodes are close to each other and have the same color, they are better connected together than with the rest of the network. You can verify this information in the Analytics > Topics panel.
Important: If you have some ideas at any stage, please, use the Project Notes field to write them down (or use "Save to Notes" button). You can use this field later to export the ideas you had during the research and generate a new text or an article from them.
Step 3: Discover the main topics, high-level ideas, and key concepts in text
Task: Zoom Out | Focus
After you generated a visualization and looked at the graph, you can use the Analytics > Topics panel to retrieve the main topics and themes in this text.
We also encourage you to use the AI: Show Categories button which uses GPT-3 AI to generate high-level ideas that describe each topical cluster.
Use this insight to enhance your interpretation of text and to get a holistic perspective.
Step 4: Zoom into the relevant topics and ideas; find excerpts in context
Task: Zoom In | Focus
Click on the topics or the nodes on the graph you find interesting to find out what other nodes / words they are connected to and to reveal the context where they appear.
This will extract the excerpts of your original text that have the highest concentration of those ideas. You can use it to read the text in a non-linear way focusing on the patterns that emerge, rather than the chronological narrative.
Additionally, you can use the AI panel to generate some ideas related to the concepts / topics you selected below.
To deselect the nodes/topics, click them again (at the top right panel or on the analytics panel) or click the undo button / reset button at the top right menu to deselect them all.
Step 5: Reveal the Structural Gaps
Task: Zoom In | Explore
Text network visualization represents the structure of your ideas. Some of the topical clusters will have structural gaps between them. This means that your particular is talking about those topics but doesn't really link them well. Bridging those gaps with new research questions and ideas is an opportunity to develop this discourse further in an innovative and interesting way.
In order to do that, go to the Analytics > Gap Insight (Blind Spots) panel. This panel shows you the structural gap: the parts of the discourse that are relevant to the discourse but could be better connected. Bridging them together will usually lead to potentially interesting and creative ideas:
You can generate these questions yourself, or also use the AI: Insight Question or AI: Bridge the Gap buttons to have GPT-3 AI do this for you.
If you develop or see an idea / research question you like, save it in the project notes:
Step 7: Generate Ideas with AI Insight Ideation Panel
Task: Zoom In | Focus
Look at the AI Insight panel and check if there are any interesting ideas proposed there, in addition to the structural gaps. You will find it at the bottom of the screen. It analyzes the structure of the discourse to discover the blind spots in your ideas. It then recommends the research questions / ideas / facts / challenges that bridge those blind spots, helping you generate new interesting insights using GPT-3 AI.
If you like what you see, you can edit the response and save it into your project notes.
Step 8: Reveal the Underlying Ideas: Beyond the Obvious
Task: Zoom Out | Explore
At some point, some of the ideas may seem too obvious to you. For example, the concepts "machine" and "learning" in a text about AI. In order to go deeper in your analysis, you can remove those concepts from the graph, to get to reveal the underlying ideas and to better understand the context behind those terms and explore the nuance.
So you can select those on the graph and remove them temporarily to see what's hiding behind those terms (see the image below). You can always get them back into the graph if you click the Undo button (next to the lock button, which permanently removes the stopwords).
Step 9: Add Your Ideas to Project Notes / Graph
Task: Zoom In | Focus
Once you get new ideas, you can:
- add a new text (you can use use the "Ideas" tag to separate your ideas from the content you generated earlier)
- add some notes about the graph in the Info menu on the left (in this case, your notes will not be visualized on the graph, they are for your internal reference only).
- click Interpret to start interpreting the data you are seeing — this will create a new graph that will be juxtaposed on top of the existing one, so you can separate your ideas from the original content.
Step 10. Reiterate Steps 1-8
Get back into the overview, use the graph to find new patterns and connections again. If at any point you like the graph you are seeing, you can always save its current version (the auto-save only works for temporary browser sessions and it doesn't save all the data).
You can also add more data to the graph from other sources. To do that, use the Add / Import panel at the bottom left.
Graph Exploration — InfraNodus Workflow Schema
Here is the ideal graph exploration workflow that you can use with your texts and files.
(please, click on the image to view it in full size)
Extra Step 11: Share your research
by downloading the graph as a PNG / SVG image, or getting a URL link and posting it on the social media using #infranodus to promote your work and ideas.
Learn how to do that in this article: How to Share Your Graphs
Extra Step 12: Compare Your Graphs or Tagged Statements.
You can also compare the graph to another graph (or the statements that belong to the different categories in your current graph) to see what the common ideas or how they are different.
Use the buttons on the left menu to do that:
Please sign in to leave a comment.