Modeling of Snow Ablation Optimization Algorithm with Deep Learning Approach for Sentiment Classification on Social Media Corpus

Abstract

Recently, Sentiment Analysis (SA) has been a tedious process in natural language processing (NLP), particularly for Social Media (SM) text, which tends to be brief, noisy, and informal. SA is a way of extracting data about an entity and automatically detecting the subjectivity of that entity. The SA aims to determine whether text created by the user conveys optimistic, adverse, or impartial feelings. The goal is to automatically classify sentiments towards specific features like products, topics, or movies, utilizing Deep Learning (DL) as an advanced technique to meet the growing demand for accurate SA. Therefore, this study proposes a new Snow Ablation Optimization with a Deep Learning for Sentiment Detection and Classification (SAODL-SDC) approach on the SM corpus. The presented SAODL-SDC approach primarily intends to recognize the class of opinions in SM data. In the SAODL-SDC technique, a multi-faceted approach begins with data preprocessing and bag of words (BoWs) feature extraction. The SAODL-SDC technique employs a Convolutional Long Short-Term Memory Autoencoder (CLSTM-AE) technique for sentiment detection. The hyperparameter tuning process using SAO is utilized to improve the effectualness of the CLSTM-AE technique. The SAODL-SDC technique is examined under Sentiment140 and the Airline's datasets. The performance validation of the SAODL-SDC approach portrayed superior accuracy values of 94.28% and 97.00% over existing techniques.