SENTIMENT ANALYSIS OF CUSTOMER REVIEWS USING MACHINE LEARNING TECHNIQUES
Keywords:Sentiment Analysis, Language Processing, CNN-LSTM Architecture, Deep Learning
Sentiment analysis is one of the most important jobs in natural language processing, and it involves extracting attitudes, thoughts, views, or judgments on a certain issue. The internet is an unstructured and rich source of information that contains a large number of text documents offering thoughts and reviews. Individual decisionmakers, businesses, and governments may all benefit from sentiment recognition. We offer a deep learning-based technique to sentiment analysis on Twitter product evaluations in this study. The proposed architecture combines CNN-LSTM architecture with TF-IDF weighted Glove word embedding. Weighted embedding layer, convolution layer (where 1-g, 2-g, and 3-g convolutions have been used), max-pooling layer, followed by LSTM, and dense layer make up the CNN-LSTM architecture. The predictive performance of various word embedding schemes (e.g., word2vec, fastText, GloVe, LDA2vec, and DOC2vec) with various weighting functions (e.g., inverse document frequency, TF-IDF, and smoothed inverse document frequency function) was evaluated in conjunction with conventional deep neural network architectures in the empirical analysis. The suggested deep learning architecture outperforms traditional deep learning approaches, according to the empirical data.