A Hybrid Approach for Sentiment Analysis in Online Product Ranking Using CNN and LSTM with Uncertain Lexical Term

Another important application of sentiment analysis involves identification of consumer attitude toward certain products, and this is particularly important in online product rankings where customers air their opinions in the form of rankings. The usual approach may fail in analysing details of human language and the nature of sentiment itself. To overcome these challenges, this paper presents CNN-LSTM method which incorporate the attributes of convolutional neural networks (CNNs) for spatial features and long short-term memory networks (LSTMs) for temporal features. CNN extracts local features in text data, while LSTM focuses on the temporal features and the combination of CNN and LSTM make this model ideal in handling sentiment analysis. Further, we present the adaptation of uncertain linguistic term sets to manage uncertainty and vagueness in sentiment expressions. The proposed methodology entails pre-coaching of the CNN-LSTM model on a large dataset of product reviews, and then the performance is compared with standard measures like accuracy, precision and F1-score. The experimental results prove that the proposed hybrid model provides better accuracy as compared to the existing traditional machine learning techniques for sentiment classification and ranking. The above approach improves the performance of the sentiment analysis by a large degree thus giving more accurate data for the e-commerce platforms.