- Tue Oct 06, 2015 5:48 pm
#30720
My name is Connor Nishijima, and I've been tracking the approximate mood of the online world since August 11th this year and displaying in color data on my desk. How is this possible? Some Twitter, some Python, and a bit of ESP8266 magic!
Here is a video:
https://www.youtube.com/watch?v=lMpsSjMVLys
Thanks to awesome developments in the tech world this last year, I'm able to use a $5 VPS to read 2,600,000 tweets a day, search for hundreds of keywords in 5 languages, and detect levels of 6 emotions:
Love (Pink), Joy (Yellow), Surprise (Purple), Anger (Red), Sadness (Blue) and Fear (Green).
The word counts found for each emotion are totaled up every minute, and the most recent data is compared to a Weighted Moving Average of the last 2 hours, to produce a percentage value. For example:
If for two hours about 1,200 tweets per minute about Love came in, but the most recent minute saw a spike of 2,400 t/min, the estimation for Love would be at 200%.
These emotional levels are then squished into 25-byte summaries and made available in plaintext via Apache2 on the VPS. Every few moments, an ESP8266 WiFi microcontroller requests this data, and parses it into a twisting color light show, thanks to the awesome 9x WS2812B leds inside. (NeoPixels as you might know them) I call this device a "moodLight". So if enough tweets like this came in:
“I am so proud of my son for getting his degree! I’m going to miss him at home.”
"Proud" and "miss him" would result in a bittersweet twist of Joy and Sadness in the moodLight. This system can be used to keep tabs on the summarized emotion of the online world, so if it fluctuates with some sour colors all the sudden - you'll know to check the news before it's even written.
Because of the wide interest shown on Reddit for my prototype, I launched a Kickstarter yesterday so people could get their OWN moodLight as well! It overfunded in just 17 hours. If you would like more information about the lamp, or to get one yourself, please visit my Kickstarter using the KickTraq image below.
[size=14pt]MoodLighting emotional analytics available HERE and HERE![/size]
- Connor
Here is a video:
https://www.youtube.com/watch?v=lMpsSjMVLys
Thanks to awesome developments in the tech world this last year, I'm able to use a $5 VPS to read 2,600,000 tweets a day, search for hundreds of keywords in 5 languages, and detect levels of 6 emotions:
Love (Pink), Joy (Yellow), Surprise (Purple), Anger (Red), Sadness (Blue) and Fear (Green).
The word counts found for each emotion are totaled up every minute, and the most recent data is compared to a Weighted Moving Average of the last 2 hours, to produce a percentage value. For example:
If for two hours about 1,200 tweets per minute about Love came in, but the most recent minute saw a spike of 2,400 t/min, the estimation for Love would be at 200%.
These emotional levels are then squished into 25-byte summaries and made available in plaintext via Apache2 on the VPS. Every few moments, an ESP8266 WiFi microcontroller requests this data, and parses it into a twisting color light show, thanks to the awesome 9x WS2812B leds inside. (NeoPixels as you might know them) I call this device a "moodLight". So if enough tweets like this came in:
“I am so proud of my son for getting his degree! I’m going to miss him at home.”
"Proud" and "miss him" would result in a bittersweet twist of Joy and Sadness in the moodLight. This system can be used to keep tabs on the summarized emotion of the online world, so if it fluctuates with some sour colors all the sudden - you'll know to check the news before it's even written.
Because of the wide interest shown on Reddit for my prototype, I launched a Kickstarter yesterday so people could get their OWN moodLight as well! It overfunded in just 17 hours. If you would like more information about the lamp, or to get one yourself, please visit my Kickstarter using the KickTraq image below.
[size=14pt]MoodLighting emotional analytics available HERE and HERE![/size]
- Connor
