Using InfraNodus, you can explore any topic using a combination of GPT-3 AI and a powerful data visualization approach based on text network analysis.
The advantage of using InfraNodus to learn about a certain topic is that it follows the ecological cognitive variability framework. This framework will guide the user to first construct a coherent discourse and will then propose them to develop the less represented concepts within to increase heterogeneity of ideas contained within. If the discourse is too dispersed, it will propose ideas to make it more interconnected.
To begin, just add any topic you're interested in and use GPT-3 AI InfraNodus module to generate 20 statements that relate to this topic. These statements will be visualized as a graph where the concepts are the nodes and their co-occurrences are the connections between them.
InfraNodus will then apply powerful graph science algorithms to detect the main topical clusters and the most influential words, so you will quickly see the most relevant ideas in relation to your query.
You will also be able to explore this idea further by clicking on the nodes or topics that you're interested in and generating more AI content for them.
If you'd like to try it with your own ideas, please, read the tutorial before or watch the video.
1. Select Explore a Topic App
Go to https://infranodus.com/apps and choose the "Explore a Topic" app.
Then add any keyword or topic you'd like to explore:
Note, that you can choose not only GPT-3 as the source of truth but also Google search results. If you choose Google, InfraNodus will get the top 20 search results for your topic. If you choose AI-Generated Facts, it will ask GPT-3 AI to generate 20 facts in relation to your query.
2. Visualize the results, check the main topics and keywords
Once the results are visualized, you will see the main topics and keywords that relate to your topic. Those are shown in the Analytics panel on the right and also directly on the graph.
As you can see, the main keywords are "meditation", "year", "vipassana" and "improve" — talking about the benefits of Vipassana meditation.
While the main topics are "Self-improvement", "Buddhism" and "Buddha's teachings" and "Meditation".
This already gives us a very good overview of the context around this topic. As GPT-3 is trained on human knowledge we can say that this topic is frequently associated with self-improvement and Buddhism.
However, we are more interested in the technical aspects of how meditation works, so let's inspect this 4th topic closer.
3. Develop a topic you find interesting
In order to explore a topic, just zoom in on it and then select it in the Analytics panel. Then generate a fact using the AI insight module and add it to the graph:
4. Select interesting concepts and link them in an interesting way
You can also click on some interesting concepts to develop them further. Just click on them on the graph and generate content that relates to them using the AI module.
Note, that you can always choose the concepts that are not connected yet if you want to get interesting results — those AI generated ideas will connect those concepts in a new way.
You don't have to click on the most influential concepts. You can also remove the most influential nodes from the graph (selecting and hiding them) to reveal the underlying contextual structure and to develop the less represented concepts and connections between them.
5. Diversify the structure of your discourse — explore less represented topics
At some point, you may end up in a situation where the structure of your discourse is biased toward a few key concepts. For instance, we added too much content about "meditation", so the structure of the discourse shows as biased toward this term.
InfraNodus will detect this change in the discourse's structure and propose to diversify it by focusing on the less represented topics. You can then develop this topic yourself or use the built in GPT-3 AI to have it generated for you:
As you can see, in this case the system propose to think of the connection between nature and working efficiency — purposefully driving us away from the topic of "meditation" but in a way that is still releveant to the main discourse.
6. Bridge the structural gaps
Once we add a few more ideas to make this discourse less focused on only one concept, we can look at the structure again and see if there are any structural gaps emerging.
Structural gaps are like blind spots between our ideas. They form when there are certain topics in the discourse that could be connected but are not yet. We can then propose a new idea that will link those topics in a new and interesting way. Usually that leads to a moment of "insight".
Structural gap information can be accessed in the Analytics > Gap Insight panel, or you can also find it yourself on the graph.
For instance, in the example above the system proposes to link two clusters: one on risk management and another on memory improvement. Connecting them with an idea or a research question will help us make this discourse more interesting. We can use our own imagination or the AI for that. For example: "Can positive effects of meditation practice on memory also help us manage risk better and be more resilient?"
You can also detect the structural gaps in the discourse if you use the "reveal non-obvious" feature which removes the most influential nodes to show the contextual structure under them.
7. Write a new text or create an outline
Finally, once you're done with the graph exploration workflow, you can create a summary of your discourse or an article outline.
To do that, click on the AI Summarize or AI Outline buttons in the Analytics > Topics panel. InfraNodus will use GPT-3 to generate an outline of this discourse based on the most important topical clusters and most influential keywords.
This helps you get the gist of the discourse without losing any important elements from your sight.
As you can see, while we started with "meditation" as a topic, we arrived to a discourse that is talking about various techniques to improve memory, concentration, and creativity — which might make it interesting to a wider audience.
To try out this approach, please, go to https://infranodus.com and choose the Explore a Topic app. Contact us if you need help with this workflow!
Please sign in to leave a comment.