An Enhanced Sentiment Analysis Model using auto Encoder Bi Directional RNN

Abstract

The proliferation of social media data had a rapid growth that made newlinesentiment analysis and opinion mining a research hotspot. Millions of twitter newline(now known as X) users express their feelings daily. Twitter sentiment analysis newlineis trending, which uses written responses to determine how a person feels about newlinea topic, which can be positive or negative. Complexity increases as the data is newlineunstructured, and this endless stream of data requires data-processing tools and newlinestrategies. The Proposed Research aims to accumulate the data, pre-process it newlineand make it compatible to apply machine learning models on it by suggesting newlineappropriate feature extraction strategies. Also this research proposes set of newlinemethodologies to deal with huge amount of social media data which is generally newlineterms as Big data. newlineAs a first model, a Tweet Analyzing Model for Cluster Set newlineOptimization with Unique Identifier Tagging (TAM-CSO-UIT) was built. This newlineapproach assigns a positive and negative value to each entry in the tweet newlinedatabase based on probability assignment using the n-gram model. To perform newlinethis effectively, the tweet dataset is considered as a sliding window of length L. newlineThe proposed model accurately analyses and classifies the tweets. newlineA second model, the Convolutional Neural Network with Optimized newlineLong Short-Term Memory Model (CNN-OLSTM), is proposed to solve the newlineincomplete, random noise that occurs in the form of different languages. The newlineproposed approach consists of three stages: pre-processing, word2vec newlineconversion, and prediction. After the pre-processing, the skip-gram model newline(SGM) based word2vec conversion is performed. The extracted vectors are then newlinegiven to the CNN-OLSTM classifier to classify a tweet as positive or negative newlinev newlinepolarity. In this, the CNN model effectively reduces the dimension of the input newlinevector using max-pooling layers and convolutional layers. Also, the LSTM newlinemodel can catch long-term dependencies between word sequences.