Social Media Sentiment Analysis Using CNN-BiLSTM

Rhea Bharal, O. V. Vamsi Krishna

Abstract: Sentiment analysis is application of natural language processing for understanding the opinions or views of public on various topics. This is also popularly known as opinion mining, the system collects, analyses and examines the sentiments present in the form of tweets. Our proposed model extracts the sentiment of the tweets and classifies them using CNN-BiLSTM which is a technique of deep learning and uses Word2Vec as word embedding layer. The Sentiment140 dataset is generated from Twitter API which consists 1.6 million tweets. BiLSTM cell state based on memory is used for tweets classification Sentiments are published on Social media in the form of texts for expressing social support, happiness, anger, friendship etc. Using deep learning approach, we will be classifying the tweets as positive or negative. CNN-BiLSTM is an effective technique as compared to others like SVM, Naive Bayes Classifier and CNN.

Keywords: CNN-BiLSTM, Word2Vec, Sentiment Analysis, Machine Learning, Deep Learning, Twitter, Natural Language Processing