How to get better text classification accuracy using deep learning and word2vec word embedding?

Question

I am trying to perform Sentiment analysis on the Urdu news dataset. I am getting an accuracy of not more than 50% can anyone suggests what\'s wrong with my code. Below is my