Open mubaris opened 6 years ago
@the-ethan-hunt That's a good idea. But I'm not sure whether it will improve the accuracy. We'll have to try it out anyway.
Right now, we are multiple features to classify, not just sentiment. If you could try a simple RNN, we can see if it will work or not.
@mubaris I guess we should try to tune the parameters and apply some boosting algorithms to improve the accuracy !!
Which ensemble algorithms should we use for improving the accuracy? AdaBoost and Gradient Boosting are good starting points IMO.
How about creating a recurrent neural network that uses LSTM sentiment analysis? The NN would generate the comments from previous comments.