The internet seems to hate word clouds. There are many sites and blogs that talk about how bad they are. I agree only insofar as they are like any data visualization. They can be good or bad depending on how they are used. If the right context is found, then any graph or data visualization can be great. 

I personally find the word within the DreamWell dream dictionary to be very useful. They seem to be a perfect data visualization for this context. So, what do they mean and how are they made?

The meaning of word clouds in the DreamWell dream dictionary

We’ll get more into this in the methodology of the word clouds, but at the core, the word clouds are showing the words associated with a specific symbol. These are the words that make the symbol unique. 

The DreamWell Dream Dictionary has defined over 1,200 dream symbols. The methodology for identifying these symbols in dreams is very simple. We simply use word search. So for the symbol of “fear”, we search for words like “fear”, “scared”, “frightened” and so on. 

With that in mind, there are three uses for the word clouds

• They help describe the symbol quickly
• They help identify new theme
• They help evaluate the quality of the symbol

The word clouds help describe the symbol

For example, here is a word cloud for the dream symbol “chasing”. It is very clear from the word cloud what is happening in the dream. People are being chased by monsters. They are running. There is a need for jumping. Just skimming the word cloud provides a very concrete demonstration of what a symbol is about. 

Here is another example from the theme of “basements”. The words here are just as useful but more surprising. There are words like “dungeon” and “cage” which are likely from nightmares about basements. There is the word “tornado” which would indicate people were seeking shelter in a basement during a tornado. 

Finally here is a word cloud for “good dreams”, which are dreams reports that contain the phrase “good dream”. You can see descriptions of how the dream was good in the words “awesome”, “pleasant”, “nice”, and “wonderful”. But you can also see some of the content of teh dream in words like “smile” and “princess”.

Word clouds help identify new themes

Here is the current word cloud for “cold”. A lot of the words make sense like “damp”, “winter”, “freezer”, or “chill”. But here is one word “sweat” which might be surprising. This is clearly from the phrase “cold sweat”. This is more of an expression than describing the experience of being cold. As a result, this phrase shouldn’t be included within the dream symbol of “cold”. This will be updated in the next release of the DreamWell Dream Dictionary, but it illustrates one use of the word cloud.

Word clouds help evaluate the quality of the symbol

While we’ve done our best to have the best data possible in the DreamWell Dream Dictionary, the quality of the different dream symbols will vary. There are any number of reasons why a dream symbol might be great or bad. 

The example above showed one reason why a dream symbol could be improved. Language is imprecise and a single word will have many meanings depending on the context. The word “loop” could indicate a time loop or it can indicate confusion in the phrase “thrown for a loop”. As a result, we tend not to create symbols that are based upon words with clear multiple meanings. 

The other reason why a symbol could be improved is because there are too few dreams that contain the signal. When this happens, the individual dreams that contain the signal overpower insights based upon the symbol across meaning dreams. 

Take the example of “nightmares”. This word cloud is very clear. All of the words clearly relate to the symbol of nightmares. This is because there are over 10,000 dreams that contain the word “nightmare”. This is a lot of data, so the individual differences have been completely smoothed over. 

The word cloud above is for the symbol “Hatman” which is a nightmare like character. These words still mostly make sense. There are words like “hat”, “cape”, and “growl” which I would expect from a nightmare type character. But there are also words that just seem confusing like “mum”, “fiancee”, or “width”. These words appear because there are less than 50 dreams with the symbol of “Hatman”, so these words are likely from individual dreams.

So the word cloud gives you a quick assessment if the symbol has enough data to draw substantial conclusions. Of course the DreamWell Dream Dictionary also provides the frequency of the dream. In this case Hatman is in 0.02% of dream while 13% of dreams are considered nightmares.

How the word clouds are made

The word clouds themselves are generated in python using the word cloud package. This is quite simple to use, but there are some tricks. 

How symbols are created in the DreamWell Dream Dictionary

The DreamWell Dream Dictionary is based upon word search. Each of our 1,200+ dream symbols has a list of words to identify it. For example, the symbol of “bathrooms” will be identified by words like “bathroom”, “washroom”, “rest room”, “shower”, and so on. 

Each symbol will also have a list of words that can be ignored. For example “cold” will be identified by the words “cold” and “freezing”, but we’ll ignore or not include any time the phrase “cold sweat” is used. 

Creating the word clouds

In theory, we simply base the word clouds upon the dreams that continue the symbol of interest. The documentation of the word cloud package will do better than me at explaining the exact math behind creating the word clouds, but these are based upon the frequency of the words. Of course some basic words like “I”, “and”, “for”, and so on are removed.

Simple Method

This is the word cloud based upon all of the words within dreams that contain the symbol of “locked doors”. This doesn’t look too bad!

However, here is the word cloud for all of the words within dreams that contain the symbol of “animals”. This seems a little strange. I see the word of “dog” and “cat”, but there are a lot of words like “around”, “see”, “time” or “feel” that don’t seem like they really belong to symbol of “animals”

This is because there are a lot of words in common across dream symbols because they all are about dreams. This in itself is interesting, but doesn’t really help provide an understanding of the specific symbol because we really want to know about the words that are unique to each dream symbol.

Based upon the difference of frequency

To do that, we’ll calculate the frequency of words in dreams with the symbol and dreams without the symbol. For example, we’ll get the frequency of all words in dreams that contain animals and the frequency of all words in dreams that do not contain animals. We then calculate the percent difference for each word between the dreams with animals and dreams without animals. 

The word clouds within the DreamWell Dream Dictionary are based upon this percent difference. This makes our word clouds a little unique. Generally word clouds report the frequency of words within text. Our word clouds report how related the specific word is to the dream theme. 

 The word cloud above shows exactly this for “animals”. It looks better! I see words like “meow”, “betta” (for betta fish), “purr” and so on. 

However, this is the word cloud for “locked doors”. This honestly seems a bit crazy. There are words like “deatheaters” (which honestly makes sense for locked doors, but it seems a bit specific), “jackalope”, “metall”, and so on. This doesn’t at all show good words for “locked doors”

This is because some words are not used that much. They are very rare within the DreamWell Dream Dictionary. The misspelling of “metall” is an example of this. It likely happened 3 times in all dreams and 10 times in dreams of locked doors. In this case, there would be greater than a 300% increase of the word “metall” in dreams of locked doors. 

Adding a filter

As a result, we institute a filter for the words to consider for the word clouds. They need to have occurred at least 200 times within the entire DreamWell Dream Dictionary. Considering that there are over 200,000 dreams, this isn’t a very strong filter. It simply allows use to remove very rare words and misspellings. 

This is the word cloud for “locked doors” that has the filter and is based upon the percent difference between locked doors and all other dreams. This shows words that are expected. 

This is the word cloud for “animal” that was generated in the same way. This shows words that make sense.

Conclusions

Overall the word clouds are based upon the difference of frequency in dreams with the symbol of interest compared to dreams without the symbol of interest. These are based only upon words that have occurred at least 200 times in the entire DreamWell Dream Dictionary. 

The aim of these word clouds is simply to give a feeling of the specific symbol. They help describe the symbol in ways that graphs cannot. They also help us improve the quality of the DreamWell Dream Dictionary. 

However, it should be noted that the analysis isn’t completed with the word cloud. It is just the start.