Over my last few articles I have analysed the raw text and relations between users in a large data set of Tweets related to Syrian Civil War.
Previous Articles in the series:
The raw text, and other textual attributes analysed shed quite some light on the discourse in the Syrian Civil War's Twitter-sphere.
But one dimensions I have yet to check in this data set was the media shared between the different users.
Using Microsoft Azure's Computer Vision API, I have automatically classified about 150,000 unique images.
Though I wish I could share those with more insight for every chart and plot, lack of personal time and the wish to share the data means I will mostly share the raw charts.
While the data collected only goes until Sep 2016, I think there is still quite a lot to learn from it.
Unique images per day
Age distribution of detected faces
Average age of detected faces per day
No extremely significant changes over time, but a slight trend towards younger faces. Pre 2016 count might be too small to draw significant conclusions.
Percentage of detected faces below the age of 16
Again we can see slight trend towards increased posting rates of underaged faces.
Percentage of images tagged "Car Bomb" per day
I was amazed that the service successfully tagged images containing car bomb in them. Would be interesting to check the dates that have significant spikes in them.
Most common celebrities automatically tagged
1 pepe detected.
If you want access to the raw data, or have any other interesting research ideas for this data-set, feel free to email me at: firstname.lastname@example.org