Wordle Visualization

Vedanshi Shah

WREN Summit 2022

Hi! I’m Vedanshi Shah

Pronouns: She/Her

Hometown: Mumbai, India

Major: Computer Science and Business Administration

​

Research interests:

• Data Visualization
• Equivariant Neural Networks
• Natural Language Processing

​

Life Mission: To have a good work-life balance :)

​

Why I joined WREN: To meet fellow researchers with shared interests and getting more involved in the research community!

M

Y

​

I

N

T

E

R

E

S

T

S

Current research

• Design spaces for visualization for communication
• With prof. Ab Mosca

​

• Hypothesis spaces for visualizations
• With prof. Ab Mosca

​

• Gauge equivariant message passing neural networks for meshes
• With prof. Robin Walters and John Park

Introduction

• WORDLE is a daily word game where users have 6 attempts to guess a 5 letter word given progress hints along the way
• The game took the world by storm in early 2022
• Our group was formed by our shared interest of the game!

Problem

• The score feature only contains a single user’s statistics
• These are nicely visualized and shareable, but not immediately comparable
• Statistics exist regarding Wordle guess distributions to see player trend over time

Goals

• Wordle is a solo player game, but a shared interest among many! There should be a way to honor that community!
• game wide statistics
• everybody’s performance records over time
• how the words have changed over time

​

Our project aims to not only capture data about the players, but also the game itself!

Design process

Ideation

Data Generation and Cleaning

Designing and Sketching

Implement in D3

Interview and Refine

Our data source

• Data Wordle tweets Kaggle dataset
• @WordleStats Twitter Account
• Twitter API V2
• Google's N-gram Word Frequency Data

Our data source

Compiled together via Kaggle

Individual Tweets

Domain tasks

• Show the difficulty of a particular date

​

• Map my performance against others

​

• Look up difficulty on particular words

​

• See if words are getting harder over time

​

• Observe if NYT made it harder

​

• Compare Wordle’s tendency to use nouns, adjectives, verbs, etc.

​

Domain tasks

Our design

Our design

LINE GRAPH:

​

Marks:

lines

Channels:

position

BAR GRAPH:

​

Marks:

lines, areas

Channels:

position, size

(length and area),

color, saturation

WORD CLOUD:

​

Marks:

areas

Channels:

size, color

Interviewing

1. Who are you and what is your community?
2. How would you describe Wordle and what is your relation to the game in your community?
3. How have you interacted with Wordle outside of your community?
4. How do you currently interact with Wordle statistics?
5. How much do you think about the way Wordle records your statistics? Do you like it?
6. Have you used a visualization tool for something language related before, such as Google’s Word Frequency Tool? If not, are you familiar with the concept of word frequency?
7. What do you look for when interacting with statistics from other users, specifically with data about performance?
8. Do you like to look at specific trends? And if so, what type of trends are you interested in?
9. A lot of what is available for Wordle is by day, by puzzle. What type of per day data would you be interested in, or data that is progressing by day?
10. Are you interested in the Wordle statistics data over a specific time period other than by day? (as a follow-up question)
11. What are you interested in learning about, specific to Wordle statistics?
12. What are you interested in learning about, specific to Wordle statistics in your community? (as a follow-up question)
13. Is there anything else you’d like us to know? (as a follow-up question)

Interviewing (important notes)

• What do you look for when interacting with statistics from other users, specifically with data about performance?
• Performance somehow standardized
• Easily comparable per case
• Numerical values
• Do you like to look at specific trends? And if so, what type of trends are you interested in?
• Better or worse over time overall
• External factors having an effect on decontextualized data
• A lot of what is available for Wordle is by day, by puzzle. What type of per day data would you be interested in, or data that is progressing by day?
• Quantification of difficulty vs how people actually did
• Data about the word itself, part of speech, building blocks, etc
• What are you interested in learning about, specific to Wordle statistics?
• Wants to graph club performance against average
• Linguists should be above average

Usability Testing

Hover Over Word

• the word being hovered over should be highlighted on both visualizations
• bar graph should also pop up an info box over the bar
• allows for a visual connection between the two different visualizations

Tasks:

​

Hover Over Word

• liked that they could view info while seeing the correlation of rarity and performance
• too much info in the tooltip and highlight was too weak to see the connection between charts

Results:

​

Usability Testing

Set Date Range

• should allow the user to limit or expand their view on the Wordle data
• allows the user to scale how much data they want to see

Tasks:

​

Set Date Range

• not fully implemented at testing
• realized that the feature would counteract the tooltip implementation
• good heads up to understand that tooltip and scrubbing did not work concurrently.

Results:

​

Usability Testing

Change Metric Toggle

• toggle to choose what to display on the word cloud vs bar graph
• adds interactivity to what metric should be displayed on each map
• observing labels, x-values, and y-values match up to the data that the toggle is pushing to display

Tasks:

​

Change Metric Toggle

• liked the ability to change between the rarity/performance, and the toggle option and the visualization was helpful and insightful.
• issues with understanding metrics and being able to see the colors representing the correct values

Results:

​

Usability Testing

Set Date Range

• Limit or expand view on data
• Scale the data

Hover Over Word

• Highlight word on both visualizations
• Info box over bar graph
• Connect 2 visualizations

Change Metric Toggle

• Toggle on display between graphs
• Interactivity to display metrics
• Observing the labels and values with the data

Tasks:

​

Set Date Range

• not fully implemented at testing
• realized that the feature would counteract the tooltip implementation
• good heads up to understand that tooltip and scrubbing did not work concurrently.

Hover Over Word

• liked that they could view info while seeing the correlation of rarity and performance
• too much info in the tooltip and highlight was too weak to see the connection between charts

Change Metric Toggle

• liked the ability to change between the rarity/performance, and the toggle option and the visualization was helpful and insightful.
• issues with understanding metrics and being able to see the colors representing the correct values,

​

Results:

​

Revised Design

More intuitive layout

​

New legend

​

Simpler color coding

​

More obvious annotation upon hovering

Domain Tasks

Show the difficulty of a particular date

​

See the relative performance of others

​

Look up difficulty on particular words

​

See if words are getting harder over time

​

Observe if NYT made it harder

​

Gauge metadata about the word itself (rarity, part of speech, definition, etc.)

Final Tasks

Show the difficulty of a particular date

​

Map my performance against others

​

Look up difficulty on particular words

​

See if words are getting harder over time

​

Observe if NYT made it harder

​

Compare Wordle’s tendency to use nouns, adjectives, verbs, etc.

Original Tasks

Domain tasks

• Show the difficulty of a particular date

​

• See the relative performance of others

​

• Look up difficulty on particular words

​

• See if words are getting harder over time

​

• Observe if NYT made it harder

​

• Gauge metadata about the word itself (rarity, part of speech, definition, etc.)

Final design

Final design

Gauge metadata about the word itself (rarity, part of speech, definition, etc.)

See the relative performance of others

Look up difficulty on particular words

See if words are getting harder over time

Observe if NYT made it harder

Final design

DONUT CHART:

​

Marks:

areas

Channels:

size, color

Demo

Thank you for listening!

Scan to view our visualization webpage

https://bit.ly/wordle-vis