Question of Interest
What is the 'favorite' number of ChatGPT?
This was the target question that was examined in this project and the results were quite fun!
As expected, based on the data that was trained, not all numbers are equally probable to occur!
To the contrary out of a sample of 10000 numbers there are 5 that occured more than 50% of the time and the most frequent of all them occured more than 20%.
So what is that number? Let me give you a hint:
Do you know about The Meaning of Life?
If you want the full story, visit the project's GitHub Page:
This repository was created to support the Honor's Track assignment of the Course:
Capstone: Retrieving, Processing, and Visualizing Data with Python, from Python for Everybody Specialization by the University of Michigan, on Coursera
The course is taught by:
Charles Russell Severance
The project was selected, designed and implemented according to the directions of the professor and in alignment with the spirit of the specialization.
Specifically, the target data source was identified, raw data was retrieved though an API, then processed to get it clean and the results were summarized and visualized.
Finally a GitHub Page was created to host and present the project.
You can either use the program to reproduce my results or to reproduce the experiment with new data. It's up to you!
First download the files with the command:
git clone https://github.com/jzstats/chatgpt-numbers-and-the-meaning-of-life
and cd into the repo.
It is required to have a valid api key that can be used to communicate with the ChatGPT's API.
Create a file called 'config.py', and put a line:
api_key = <the-api-key-of-the-user\>
This file will be sourced by 'main.py' when it tries to retrieve the raw data.
(it was not included in the repo to hide the API key used for the project, which is my personal one!)
In order to reproduce my results run the 'main.py' script.
When it asks for the number of desired observations, input 0 and you are practically done!
Next you can explore the top n results and may compare them with mine.
Finally to produce the wordcloud, run the script 'visuallize_results.py' and open in a the browser the file 'wordcloud.htm'
If you wish to reproduce the same experiment but with new data you should first delete the 'data/rawdb.sqlite' DB.
Then run the script 'main.py'.
When it asks for the number of desired observations, input the number of observations you want to collect. The data retrieving process is quit slow and may take from minutes to hours depending on the sample.
THINGS MAY BREAK in the data processing step, as the cleaning and extraction is sensitive to the data I had collected and it is possible that you may need to modify the script 'process_the_raw_data.py' to fit your observations.
Then you can explore your results and may compare them with mine.
Finally to produce the wordcloud, run the script 'visuallize_results.py' and open in a the browser the file 'wordcloud.htm'
The scripts and files used for the visualization, were taken almost as is from contents of the courses and specifically correspond to the files:
- 'visualize_results.py' to 'code3/gmane/gword.py'
- 'wordcloud.js' to 'code3/gmane/gword.js'
- 'wordcloud.htm' to 'code3/gmane/gword.htm'
And the JavaScript Libraries as is:
- d3.v2.js
- d3.layout.cloud.js
from the repo:
Intentionally, to conform with what the openai's API demands.