Dream Analysis and Interpretation with GPT AI and Text Network Visualization

Posted Sunday, March 26, 2023 by Dmitry Paranyushkin

How can AI be useful for analyzing dreams? At a first glance, this may seem like a weird idea. However, advanced data science and AI tools can provide very interesting insights into our subconcious patterns and reveal interesting patterns in our psyche. Moreover, as we become aware of our dreams and the patterns that tend to reoccur on a regular basis, we can also foray into lucid dreaming: the ability to become conscious of one's dreams and actively guide its narrative.

In this article, I will demonstrate how InfraNodus and the built-in data science and network visualization tools can be used for dream analysis. As an example, I will use a journal of my own dreams recorded with InfraNodus between 2015 and 2022. I have to admit, it's highly personal, but I thought it would be interesting to use a real-life example. This approach can be interesting not only for introspection but also for psychology as the content of our dreams can reveal hidden patterns in the subconsious. Using advanced data science methods and AI can help us enhance our undertanding of the psyche and lead to very interesting insights.

 

On the image above, you can see the top concepts that reoccur in my dream logs visualized using InfraNodus. I then used the structural gap detection algorithm that identifies the blind spots: distinct recurring topics that almost never occur together. Then I use the GPT AI to generate a question that would link these two topics together, providing me with an interesting insight of how I could rewire my own unconscious to connect disparate images that tend to reoccur in my dreams.

It can be seen, that two of the topics I was dreaming about were "creative expression" and "human interaction". InfraNodus proposes me to think how they could be connected. I like to do that myself (there's plenty of ideas to come up with), but out of curiousity I like using GPT AI to generate some ideas for me. Here's what it came up with: "What impact does the perception of fatherhood have on the body image of women when they explore lost emotions through music?". This is an interesting idea, because it makes me think of how I could explore an embodiment of "motherhood" when I listen to some emotional music (in reality or in dreams).

This is one of the final results of the analysis, I will show how it could be performed step by step below.

 
 

Step 1: Getting a Log of Your Dreams into InfraNodus

The first step is to get the dreams into InfraNodus. The easiest way is to simply copy and paste your dream log into a new context graph in InfraNodus, visualize and analyze them all at once. If you would like, however, to have the ability to compare your dreams one by one, to compare them, to trace how they evolved over time, and to search through them, you can also use InfraNodus as a dream log for a period of time and save every dream into a separate context graph.

Step 2: Becoming Aware of Recurrent Dream Patterns

Visualizing the contents of dreams makes it possible to not only analyze their content, but to also get an overview of recurrent images and ideas that our unconscious is dealing with. Research has shown that we dream during the REM (rapid eye movement) stage of sleep when our brain is rapidly reconnecting different neuronal pathways in order to try out alternative versions of reality and to resolve some of the issues or problems we have during the day. Getting an insight into the structure of those occurrences can help us learn more about ourselves and to reveal some deeper truths that we are trying to resolve or recreate on a deeper level.

Once you add your dreams to InfraNodus, they will be visualized as a graph. The concepts that appear in your dreams will be shown as the nodes and co-occurrences of those concepts are shown as connections. (You can read more about the algorithm used on the how InfraNodus works page). This representation will allow you to see the most influential ideas (concepts that tend to connect different topics together and serve as the pathways of meaning in your dream-making) as well as reveal the groups of concepts that tend to reoccur together in the same context.

In our example, we can see that the most prominent concepts (shown bigger on the visualization and also highlighted in the Analytics panel) are:

• "people",
• "house"
• "start"
• "time"
• "place"

These are the main protagonists of my dreams: places and people. Note, these are not necessarily the most frequently used concepts, but the concepts that connect the different ideas in my dreams most often. There is a difference, because we use betwenness centrality measure to calculate the importance of concepts and this is different from frequency although it does correlate with it.

We can also see that the main topical clusters (shown with distinct colors on the visualization and in the Analytics panel) are:

• • house place room
• • stay found guy
• • thought good felt
• • father woman tell

These are the recurrent patterns in my dreams: concepts that reoccur in the same context more often. As we can see, I'm dreaming about various locations, people, feelings and thoughts, and my parents.

At this point, I can use the built-in GPT AI to generate the names for those topical clusters above, so I can have a general understanding of the main topics that recur in my dreams:

What I find interesting here and what I like is that I often dream about creative expression and writing. This happens at the backdrop of (hopefully) beautiful landscapes, but also at urban landscapes. I'm often on some kind of road trip and there's also the notion of home or some space that tends to come up frequently. Finally, there's plenty of decision-making processes and human interaction in my dreams.

All this is very useful information, because I get insights about my subconscious dreaming and what my unconscious is trying to deal with while I'm in the state of sleep. Moveover, being aware of our dream patterns makes it more likely that I'm going to recognize them again while I'm dreaming, increasing the chances of a lucid dream.

Now that articicial intelligence identified recurrent topics in my subsconscious, let's move one step deeper and see what is hiding underneath.

Step 2: Reveal Underlying Ideas

The next step is to reveal underlying ideas: recurring concepts that come up in dreams but are not very visible at the first sight.

InfraNodus has a very useful feature for this: you can select the top influential concepts and hide them from the graph to highlight the context around them. What happens as a result is that other, less visible ideas, become more prominent and you can better understand the nuance and specifics of dreams. In order to do that, go to Analytics > Top Influential Nodes > Reveal Underlying ideas:

As it can be seen, the underlying top four topics are much more specific:

• • Lake Beauty
• • Room Place
• • Waiting End
• • Road Driving

We can interpret these topics by saying that I often dream about nature and especially beautiful lakes (and water), but also spaces and rooms, and I either wait for something or travel somewhere.

The 3 remaining topics are:

• • Car Thing
• • Father Woman
• • Water Find

Which make the interpretation more precise by adding that there's recurrence of cars (while driving) and prominence of water (and also searching for something in relation to water - flux). There is also an interesting topical cluster that contains information about my father (and to a lesser extend — mother) and a woman — these are the generic protagonists I dreamt about a lot.

One could then go deeper into each topic and discover if there are any interesting patterns in the subconsious that relate to certain ideas.

Step 3: Diving Into Interpretations

In order to dive into a certain recurrent notion in dreaming, we can select a word on the graph and InfraNodus will show us all the statements that contain this word.

For example, I'm interested in which context the notions of "water" and "river" appear in my dreams. I'm going to select this word on the graph and InfraNodus will show me the concepts that appear in the same context. I will also be able to see the statements (logs of my dreams) where this concept appears:

As we can see, InfraNodus shows the statements where the both notions appear on the left (from different dreams). The Analytics > Relations panel on the right shows the other concepts both "water" and "river" are connected to (they are also shown on the graph): "lake", "dream", "end", "berlin", "escape". I can interpret this information by saying that water and rivers (and lakes) appear in my dreams with a notion of escape or something ending.

It is important to emphasize this step: we just derived an interpretation of symbols based on the context of the dreams themselves. We did not look into a book that tells us how we should interpret water, but, rather, just observed what is the context where the notion of "water" appears in our dreams. Of course, we can also add to our subjective interpretation a more general interpretation as well, which would take archetypes and collective experience into account.

Morever, we can use the built-in GPT artificial intelligence tool to generate a question to our subconsious. We can do that by clicking the AI: Generate Content in the Analytics > Relations panel. It will generate a question connecting all the ideas that belong to the "water"/"river" cluster. In our case, it's asking about about the effects of escaping to rivers or lakes on the unconscious. Perhaps, an idea to think about on a journey to nature.

Step 4: Navigating Through Dreams

We can use the interactive text network visualization to navigate into dreams, which have the images and symbols we're interested in.

For example, if we click on "water" and "river", we can then choose to load the dreams that have the highest concentration of these images. These will be shown as square nodes on the graph. Double click on a graph to jump into it and to see in which context the term is used in a particular dream to reveal the context:

In the example above, we jumped into a specific dream, from 8 years ago, that talks about "water" and "river". Interestingly, it happens in the context of exploration, rather than escape, which provides a more nuanced view of the context where these symbols appear. After all, water does not always mean running away from something, it might also mean exploring something. Perhaps, running away is, in fact, exploring.

Step 5: Ideating through Dreams

We can also use this dream text analysis to generate new interesting ideas to stimulate our subconscious. One of the possible approaches is to use the "structural gap" feature of InfraNodus. It identifies two distinct topics that could be better connected and proposes to bridge them together with a question or an idea. For instance, in the example below we use the original graph where InfraNodus identified two topics:

• • Human Interaction
• • Creative Expression

It then proposes to think of a link between them. We recommend to try to bridge the gap using the brain, however, it can also be done using the built-in GPT AI. For instance: "What if I start organizing events where I would invite my friends to create something together?".

Interestingly, the way it works is through identifying two recurrent topics that seem to come up during sleep in the subconscious. As they are distinct from one another, they probably rarely connect. However, their presence in dreams indicates that they are somewhat important. Thinking of a connection between them helps us connect different parts of our subconscious, which may hopefully lead to some really interesting creative results.

 

Please, subscribe to our newsletter to learn when we publish more tutorials on using data science and AI for introspection.

---

Try It Yourself

You can analyze your own dreams with InfraNodus using this approach. Just sign up for an account or log in if you already have one:

Sign Up     Log In