cjhutto
Vader does not predict correctly the sentiment of some emojis.
Vader does not predict correctly the sentiment of some emojis e.g. 🤬 , ❤️
This might relate to my issue #99. Have you checked whether my hot fix works for you?
Sorry to came late to the party. After finding some sources for another problem, I found this interesting take on emojis on this kaggle website: Twitter Sentiment Analysis
According to the creators of the dataset:
"Our approach was unique because our training data was automatically created, as opposed to having humans manual annotate tweets. In our approach, we assume that any tweet with positive emoticons, like :), were positive, and tweets with negative emoticons, like :(, were negative. We used the Twitter Search API to collect these tweets by using keyword search"
citation: Go, A., Bhayani, R. and Huang, L., 2009. Twitter sentiment classification using distant supervision. CS224N Project Report, Stanford, 1(2009), p.12.
I actually was testing this in my vader implementation using emojis such as 😀😂 as positive and emojis such as 😠😡 as negative. Obviously, one would hope that such emojis wouldn't be used sarcastically in a way that it tampers with the sentiment detection of a certain phrase. However, it's pretty interesting that a simple solution like this would result in a pretty accurate dataset for Sentiment Analysis.
I would suggest to the creators of VADER that a very good idea for the emojis would be to be treated like this, you can just simply have a table/array with the emojis and when a emoji is detected, you just convert it to the specific sentiment and that's it.
