ETL to QE, Update 12, Presentation at Meetup
Date: 2023-10-18
See Discord Binding for project context
Today I attend the Creative Code Toronto meetup where I met plenty of interesting people coming from a array of varied background all working on unique projects. When the organizer requested if anyone also wanted to present their projects I put my hand up and was selected.
I ended up connecting my MacBook to the display and showcasing the graph's of my project which you can find here. Once in front of an audience I realized how much context my jupyter notebook was lacking.
I had to explain my analysis orally
When presenting I just walked up in front of everyone, spent 2 min explaining the project context, and started showing off the graphs. Upon opening up the "Message Counts" graph I realized just showing a cool chart to people is sort of useless. I had to explain the results. In this case I talked about how certain discord servers have serious bot integration sometimes having more messages from bots than actual people. That is an interesting insight people might not have known about before.
When I got to the "Accounts, Channel, and Message Counts" graph which has 3 independent vertical axis I realized I should have just produced more specific charts because there was too much going on to produce any insights without mentally segmenting the data.
When I got to the normalized graph's at the bottom of the Jupyter Notebook I explained how there is no monthly regular activity in any of these discord servers. They tend to peak at one point with the following and preceding months having much less activity. I realized that this analysis is pretty apt but I had not explained it anywhere in the jupyter notebook.
Feedback recommending look for key words
After all the presenters at Creative Code Toronto we all had some time to mingle and some of the other guests at the meetup came over to me to talk about my project. One person recommended that I look for key words such as stock tickers or project names within the discord data. I explained that I was planning on producing an interface that can return search results. The guest then explained to me that I can graph how many times certain key words are mentioned in each guild. This was a very valuable piece of feedback I plan on implementing.
Message Interrogation Interface
When discussing my project with other guests at the Creative Code Toronto meetup I realized I had an entire dimension of data I had not used in my analysis yet. When indexing the Discord Messages from JSON to SQL I have a column content_length in table messages_t which contains how long each discord messages is. So far I have queried the longest messages from guilds and put them into a dataframe and read a few but I have not analyzed what they were about. Actually I can now query what users have the longest average message and that would signal who is worth trying to talk to within these discord servers.