An Efficient Neural Network Model for Sentiment Analysis using Bi-LSTM and CNN

These days, internet users are extremely keen in expressing their thoughts on a variety of goods or products in social networking sites, which has led to an increase in the amount of human generated content found on the web. Analysis of user sentiment on social media platforms like Twitter has emerged as a highly essential and difficult job in recent years. A nontraditional technique is required in order to complete the job of sentiment analysis in such a scenario due to the features of the data being used. In addition, analysis of sentiment on social media platforms is a basic topic that has a wide range of fascinating applications. The majority of the currently available algorithms for classifying sentiment on social media determine the polarity of sentiment solely based on the textual content of posts and ignore additional information available mainly on the following platforms. This In this work recommends an architecture for combining CNN and Bi-LSTM models into one another. The implementation of BILCNN which is a combined approach includes a word embedding concept that transforms tweets into numerical form, a CNN layer that accepts attribute embedding as source and generates features with a smaller criterion, and a Bi-LSTM model that takes input again from CNN layer and generates classification results. GloVe and Word2Vec were both used in their own right in order to determine the effect that the word embedding response had on the model that was suggested. The datasets for Yelp and Tweets were obtained, and BILCNN was applied to them. The findings suggest that the suggested system is capable of providing suggestions by a precision of 94.8 percent, offering consumers a satisfactory performance in their recommendations.