In the past few years, there has been an enormousincrease in the textual data from social media (Twitter, Google,Quora, Reddit) over the internet. Sometimes the text is opinion-ated, and authors have different emotions behind that, whichrefers to sarcasm and identifies these comments as Sarcasmdetection. Sarcasm is often used to mock people, generally seenas using positive words to express negative emotions. We aimto analyse the sarcastic and non-sarcastic comments on theReddit website. In this project, we have applied a wide rangeof classification algorithms on different feature sets to check theaccuracy of every situation to classify whether the statement issarcasm or not. Our experimental results show a Gated RecurrentUnit model gives an accuracy of 73.73%, which is the highestamong all attempted approaches, while high loss indicates roomfor improvement.
- Josip Cerovac
- Siddarth Shantinath Patil
- Vipin Bairagi
- Zhengxing Li
The report for this assignment is availale here.
A short presentation video regarding this assignment is available here.