The Gram = Instagram
Social media are user-generated-content (UGC) and this makes them rather interesting in understanding the human condition, surfing internet images of cats and acknowledging the role of distraction in our lives.
Analysing images and text on social media is becoming increasingly tricky and a game of cat and mouse. Platforms change ownership, update APIs* (how applications talk to each other), and enable spontaneous new forms of interaction simply by being there, hello #hashtag.
Recently, I've fallen down the rabbit hole of learning (very basic) natural language processing (NLP) using Python. What is interesting is how quickly you can pull in a corpus of text (basically a sandbox of text signifiers and classifications) to understand associations from social media content. Now the system is not foolproof, so the reliability and validity of such results are at stake, however, this does allow tentative review of the kinds of content being shared. My investigations analyse some of the most popular #hashtags on 'The Gram'
#sunset, #style, #money, #healthyfood, #photography, #WTF, #brand, #recipeoftheday #bekind, #travel, #fitness, #cats (they get everywhere).
Close associations to the above follow a pattern of overt and tedious marketing content (yawn), aspirational 'stuff' designed to make you feel inadequate, the acknowledgement of the rougher edges of social media and need to #bekind, along with a range of emoticons and slang. Lovely stuff.
smsdictionary (I need this one, I did not know an aubergine meant that...)
Some bright sparks (Asghar and friends) have pulled together a partial list of slangs with their sentiment class:
Coolio - Cool - Positive
gr8 - Great - Positive
Xoxo - Hugs and kisses - Positive
Air - Alright - Positive
Happs- Happy - Positive
Smh - Shaking my head - Negative
Damn - Disbelief/condemn - Negative
Hehehe - Laughing - Negative
Notta - Not - Negative
Chale - Disagreement/disproval - Negative
Gonna - Want to go - Neutral
Haha - Laughing - Neutral
RoflRofl - rolling on floor laughing - Neutral
Wanna - Want to - Neutral
The main challenge is context. 'Hehehe' vs 'haha' are very context driven, as is 'smh' - ah the joy of sarcasm. I endevour to continue to build my own directory of social media slanguage, picking up on ways we modify our understanding and enable contextual use of algorithms to classify and unpack meaning.
*application programming interface (API)
Asghar MZ, Kundi FM, Ahmad S, Khan A, Khan F. T‐SAF: Twitter sentiment analysis framework using a hybrid classification scheme. Expert Systems. 2018 Feb;35(1):e12233.