Dependencies

simple-image-download
instaloader

Data Collection

• Google - Data is collected form google search API using keywords for each of the 6 pillars
  
• Instagram - We use instaloader library(https://pypi.org/project/instaloader/)
  usage - instaloader profile [instagram page handle e.g. memedefsg]
  This creates a folder with the handle name and downloads visuals
  

	#Visuals
Google search	2,125
Instagram	33,954
Total	36,079

Instagram handles Used

| #mindefmemes | | @kmfst_ | | @memedefsg | | @sgagsg | | #sgmeme | | #sgmemes | | #singaporememes | | @yeolo.sg | | @mndsg_ (memes n dreams singapore) | | @nus_memes | | @smumemes | | @pioneer_university | | @changicollege | | @sgsocialworkmemes | | @mas_complaints | | @sit.tum.memes |

Dataset Annotation

Dataset	#Visuals	classes
Non Memes	1,711
SG Memes	2,893
Non-SG Memes	2,408
Pillars - Stance	2,513	[Social,Economic,Psychological,Military,Civil,Digital,Others]-[Supportive,Neutral,Against]
Tags	2,513
Text	7,012

Dataset is available here
This dataset should only be used for non-commercial research purposes. Should you find any meme in violation of copyright, please notify us and we will take it down.

Annotation is available in:

• report/annotations_raw.csv - contains raw annotation by all annotators
• report/annotation.json -
  
  • contains labels based on agreement between annotators
    
  • Tags are combined excluding duplicate tags and removing stop words

Annotation stats (Krippendorf's alpha):

Types	Pillars	Stances
0.65	0.55	0.21

To cite:

To be Updated