An Experimental Framework of Bangla Text Classification for Analyzing Sentiment Applying CNN & BiLSTM

Analysis of sentiment is the utilization of NLP, which works with data to analyze emotions or thoughts that may be useful in many aspects. Sentiment analysis with Bangla text has been a challenging task as only several research on it. As a decision-maker, emotion extrication catches customer perceptions and helps in social behavior. In this paper, we have explained a new way of analyzing sentiment applying text classification on the Bangla dataset utilizing the CNN-BiLSTM framework. This classification model presents the text as either positive or negative. In our model, we have used a dataset on Bangla news comments, consisting of Bangla posts for developing a deep learning model. To achieve accurate analytical results, we have prepossessed our dataset involving language translation, stop word elimination, tokenization, and high-frequency word extraction. Then to generate the sentiment analysis model, layers of CNN and LSTM are designed in such a way that we have successfully achieved our high accuracy.