a brief study into learning how to categorize data in sketchnotes form

As practice into forms of data visualization, I took my notes over data in sketchnotes form, focusing on what was most important and emphasizing facts with illustrations and hand lettering.

Through note taking and in-class study, I developed my understanding of what data was, how the author's worldview can impact it, the different levels of data (entities, relationships, and atributes), and how data can be wielded to empower/harm.

brainstorming ideas to try and record my weird and wild life

With the introduction of this project, I already had plans of trying to record something a bit more out of the ordinary. At first, I had considered two ideas.

- "How could I document the amount of time I spend using my legs?"

- "How could I explore my habit of data hoarding through open tabs on my desktop?"

Though these ideas did sound interesting, I felt like their scope was a bit limited and in tune to my professor's feedback, could be weirder. And so, I went back to the drawing board, immediately opened up Discord instead to check on a notification, and there I saw it.

Much of my time has been spent by shitposting on random Discord servers, specifically with this emote. Several years ago a friend had introduced me to this emote. :naruhodo: It originated from a Bkub comic from 2017, to a Russian Telegram sticker pack, to a Twitch emote, to a friend’s discord server server - all under different names), and I proceeded to run with this joke until the present day.

“Naruhodo” is a japanese word generally meaning “I see”, “I agree” or “I understand”, generally used in a more casual environment as it includes feelings of admiration.

Similar to how it is used in Pop Team Epic, how I use it is almost completely opposite of what I mean. I will use the emote for any context, whether I'm agreeing, disagreeing, chastising, waiting, etc. In addition, there are many variations of the emote made by myself and others that I use in tandem with the original, with their use sometimes making even less sense. Because of this, I thought it would be interesting to track how I use :naruhodo: in order to see what the emote truly means to me.

how will i collect my data?

date	time	platform	community	variation	context	purpose	placement
ordinal	ordinal	nominal/text	nominal/text	nominal/text	nominal/text	nominal/text	nominal/text

The emote is primarily used online on Discord and Twitter so timestamps and contexts are recorded when I use them. I plan on collecting all of the data at the end of the day as to not disturb conversation flow in cause I use the emote multiple times in a conversation. I have used it in real life on rare occasions, so for those times I will record the use immediately.

I had brought up the history of the emote during discussion with my peers and professor, which lead to interest on the history of the emote even beyond my usage. Tracking the evolution of :naruhodo: across a longer time frame seemed interesting as I feel like it could also reveal more about my messaging habits in general. With this, I decided to shift my original collection protocol to focus on the one community I talk the most in, over the period of a year, taking random samples from each month as using all 1,000+ instances would lead to a very bloated visual.

date	time	variation	pre/post messages	purpose	placement
ordinal	ordinal	nominal/text	??? image/text?	nominal/text	nominal/text

My data collection process would be mostly the same, although I would be choosing 25 messages at random from previous months, spanning from August 2022 to August 2023 to take a full year of :naruhodo: use into perspective.

learning more on good practices for visuals

Going through the mini-lecture, I started to consider what kind of visualization I wanted to do. Considering the amount of data points I had, it would probably be best to use color as my indicator between different types of data, since the scale of my visualization would lead to images being nigh unidentifiable. Again, with so many points, I decided that a dot plot would be the best way to show my data.

welcome to hell, enjoy your stay

I initially started with manually hand typing pre/post messages, but I realized after the first month of messages...

This would take years for me to complete.
Then I started taking screenshots instead.
I realized this would also take years
And also heavily bloat my visualization.
Well damn, what now?

I ended up finding an extension called Discrubber that allowed me to search my use of :naruhodo: spanning back to 2021, and download all of the message timestamps and contents. Minor editing to the data had to be done to convert it to my timezone. (and had to consider daylights savings time)

I wanted to include some information on my visualization about how often messages were chained together. What is a chain you might ask?...

aChain

- Within the same conversation

- No more than 10 min between messages

bChain

- ~3 min between messages

Does not account for...

- Stickers

- Animated Emotes

- Variations with 1 appearance

If I had the knowledge, I might have been able to automate the process, but instead I ended up manually reviewing all 1.2k messages. (took a very LONG time, several hours). Part of me thought that I might be pushing the scope of my project, but the other part of me was already in the deep end. There was no backing down now, I would process every message.

what is this even about???

Next was determining how to classify message purpose. Since this information would be purely subjective, I would also have to spend several hours manually processing data, so I worked on finding ways to make the process more efficient. I started by going through every message in July 2023 and describing every message however I saw fit. I then reviewed the descriptions to try and find common themes to categorize messages like “bullying” and “deflecting”. I started to reclassify, but I realized that the words were still too long and I still had a few too many classes. And that’s when it dawned on me.

It’s time for a fucking alignment chart baby.

Through careful introspection the :naruhodo: alignment chart was created to define my messages on a Malice x Activity scale.Each spot on the chart had their rating like Active Benevolence that could be shortened to the first letters (AB) for faster processing. The phrase word, “Heh, Nice” helped define the key theme of the spot, while the sample message and keywords helped fill in the spots that the phrase might have missed.

And with the creation of the :naruhodo: chart, I delved back into the Discord depths and began logging in data once more. My new data collection protocol proved to be very effective and I ended up finishing all of the messages in about 5 hours.

Data Collection Spreadsheet.

experimenting with various programs

Over the weekend, I played around with Rawgraphs and Observable to create some data visualizations. What I realized was that something about my data was off, both sites couldn’t read my entries correctly if I wanted to use specific visualizations.

Also, Observable’s documentation is dogshit and made me want to cry.

It is very funny to see all of those points and it definitely made me feel validated about recording the full 1032 messages, finding that interesting data about myself was already emerging. I would spend a few more days working with Observable and while I found that the information that they revealed was interesting, I thought it would be confusing for people not in the know.

it's time to draft

At this point, I was determined to find a way to include the full 1000+ messages in my final visualization. This would be a tough task as many of the observable plots I had made ended up extremely cluttered, with no way of appreciating the full magnitude of 1000+ points. I had different ideas like basic bar graphs and bubble charts, but the two that stood out to me the most were

- A physical visualization that had each message printed on small strips of colored paper contained inside a large jar

- A digital visualization that spewed out a shit ton of :naruhodo: as a physics simulation

Although my original intent was to create an interactive digital experience to view my message contents, something about a physical manifestation of my messages enthused me greatly. I felt that the mass of 1000 messages can’t really be felt in a 2D visual experience as it is all contained with a screen. Being able to physically interact with each message would help show my absolute obsession and overuse of :naruhodo:. I also felt as if the physics simulation was a bit of a surface level interpretation of my overuse of the emote, so with this in mind, I pivoted and started developing the physical visualization

honing in on an idea

In discussion with my peers, it was agreed upon that the physical visualization was a more compelling idea. It was suggested to separate the messages out into different containers in order to show the magnitude of each alignment. Through this, I had decided on using acrylic cubes, so that they could form a grid and take the shape of the alignment chart itself. The idea was to print and attach the alignment chart to the back of each cube so that the information about each alignment could be included on the physical visualization so that when the cubes are all set together, it would form the alignment chart. To help tie the visualization back to :naruhodo:, I changed the strips of paper to laser cut cardboard :naruhodo: silhouettes. At this point, I had thought that the repeated alignment classification of color and location was unnecessary, so had decided at the time to leave the cardboard neutral colored.

After another round of critique, I had run by the idea of my physical visualization to my professor for feedback, and was asked to try and encode more data into the visualization with color. The problem was that most of the other data I had was either fairly linear and boring to visualize or overwhelmingly skewed in one direction that the visualization would not be able to represent due to physical constraints. At the same time, I was questioning if users of the visualization would be able to interact with the interior pieces, or should I keep the tokens contained within the cubes. This brought on the idea of having the visualization serve as not only a visualization of my messaging personality, but also an interactive experience reproducing my time categorizing all of the messages with the alignment chart.

Each token would once again be assigned a color according to their alignment, and I would ask users of the visualization to sort the data out themselves in order to discover the data. Even without the tokens sorted through, my hope is that the weight of the 1000+ messages is felt through a giant pile of :naruhodo: tokens, and that the uneven distribution of colors would be noticeable in the pile.

The final iteration of the visualization before the creation process would involve:

Data + aesthetic mappings:

- Colors + position in grid to show alignment of message

- One message = one physical token

Geometric Objects:

- One token represents a naruhodo

- Acrylic cubes are the squares of the alignment chart

Scales:

- X - Activity

- Y - Malice

it's time

And so, with everything ready, I started to produce the final visualization.

And it immediately started awfully. My cubes got lost in transit and I had to rush to order more. Oh well.

I went ahead and took the time to run material and size tests. I had the choice between 1/8th in. plywood or 1/16 in. cardboard, and while the plywood had a more premium feel, the cardboard was a much more economical and environmentally friendly option. I also tested sizes ranging from 1/4 in., 1/2 in., and ¾ in. widths, ultimately settling on 3/4ths by doing some volume calculations and accounting for the random jumbling of tokens to take up extra space.

Doing some calculations, with one token taking roughly 15 seconds, it would take 4.3 hours to cut all the tokens.

4.3 hours

Yeah, no - We’re going to optimize the crap of this. I solved some issues with lines doubling through printing with InkScape, and manipulated some of my strokes to help with efficiency. I ended up with an estimate of about an hour which I was pretty fine with. With everything ready, I hit start…

It only took 30 minutes.

My estimate was extremely off. It worked out in the end so I guess it was a win?

Anyways, I went ahead and sorted all of my tokens into their respective categories so that they were ready to be dyed, making sure to inhale a dangerous amount of fumes during the process.

Next came the dyeing process. I thought it would be pretty easy - mix some food colors with water, let the cardboard absorb the dye, bing bang boom done.

It took me 2 hours. Completely the opposite of what happened with cutting.

💀

I had to spend a lot of time remixing colors since my mixes ended up being either too strong or too weak, but eventually I got every piece dyed. However, some sets ended up being too close in color so I had to redye them the next day.

At last, after several hours, I was done.

At the same time, I was working on the alignment chart graphics for the cubes. For the most part, it would be the same information from the original alignment chart but in the style of the discord client.

I also added the counts for each alignment chart value, just so that information is easily viewable since there aren't hover options in real life like there are on the Observable interactive graphs.

These would be put inside the cubes so that they could be displayed out the back. The original plan was to have a large :naruhodo: displayed on the front side, since I didn’t want to obscure any information, but after talking with my professor, I decided to just repeat the information and shift it slightly higher so that it could be mostly viewed over the piles of :naruhodo:.

I initially wanted to include a sample message on the backs of the cubes so that people with no knowledge of my messaging habits could discern what my alignment chart meant, but there was no good way for me to incorporate it into the cube without it getting hard to handle. Instead, I compromised by creating a separate reference sheet as a complementary piece to the main visualization.

my hands hurt

Probably could have had a better idea of what i wanted to do from the beginning instead of shifting so much to save time Chains were interesting but ended up not being implemented in the final, maybe spending the time to physically chain them together in a V2 that’s based more on my use over time?

1. Perspectives of other users is missing, would be interesting to see how much my interpretation differs from the average reaction to the message
2. All of the data collected out of context could definitely be used to portray me in a very VERY bad light, as shown by the sample messages.

A lot of the data I had collected manually, such as chains, message contents, and message placement ended up not being used in my final visualization. While that content was interesting to track and learn about, I wish I had planned out my protocol better earlier so that I didn't end up spending time manually recording so much data.