Deep Learning (DL) schemes offer solutions to a wide variety of problems. Mainly, two kinds of Neural Networks (NNs) namely, Convolutional Neural Networks (CNN) and Bi-directional Long Short Term Memory (LSTM) networks are used. CNN+Bi-a LSTM based DL scheme is used along with pre-trained scheme for automatically determining features to explore sentiments and categorise reviews as well as opinions based on positive or negative polarities. The proposed model offers better performance on standard datasets. The performance of proposed mechanism is compared with standard DL-based schemes. CNN+Bi-LSTM scheme is proposed for Sentiment Analysis (SA) of Twitter data. The performance of different Word Embedding (WE) systems like Word2Vec and global vectors (GloVe) for word representation are compared based on best scoring values. The proposed scheme offers 93% accuracy for categorizing tweets into negative as well as positive cases on benchmark dataset. The scheme uses LSTM-based NN with limited factors offering modest outcomes and outdid other ML schemes in terms of Accuracy, Precision, Recall and F1-Score.

Sentiment Analysis, Deep Learning, Convolutional Neural Network (CNN), Bi-directional Long Short Term Memory (LSTM), word embedding, Twitter data