SENTIMENT CLASSIFICATION USING HYBRID TEXTBLOB BI-LSTM DEEP LEARNING MODEL

Sentiment analysis techniques are used to classify tweets as positive or negative. While traditional machine learning methods often struggle with low performance, deep learning models Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory (Bi-LSTM) have shown significant improvements. In this method a novel hybrid TextBlob-Bi-LSTM method is proposed for sentiment analysis and the tweets are classified here based on sentiments extracted by TextBlob using a deep learning Bi-LSTM classifier. TextBlob is a lexicon based tool which is interpretable and easy to use but lack adaptability. Bi-LSTM is used to improve the context learning through bidirectional layers but struggles with parallel feature extraction. Combining these two in a hybrid model allows us to exploit the semantic understanding of TextBlob and the contextual learning power of Bi-LSTM. This proposed hybrid model achieves the accuracy of 89.3 and this hybrid model typically performs 2 5 better in accuracy compared to LSTM depending on the common dataset.