Sentiment Classification in Bengali Linguistics Using Directed Machine Learning Techniques

Abstract

Sentimental evaluating is part of NLP research (SA). This dataset was created by scraping information from social media. The contents are also carefully categorized into positive and negative categories. Another name for this is polarity categorization. Emojis are becoming more widely utilized in written communication to express emotions or to repeat statements. Prior AI systems just looked at the arrangement of text, emojis, or pictures, with emojis with text continuously being disregarded, resulting in a variety of feelings being missed. A composite process technique of the "Pipeline" class is used to extract features and train the dataset, which incorporates Count-Vectorizer, transformer, and machine learning classifiers. In terms of accuracy, RF surpassed the other five classifiers. Even though LSVM has the lowest accuracy, it is also nevertheless gives excellent results. However, for current and critical linguistic data, this work has produced superior results, suggesting that adequate feature extraction was used to develop the model. In case of using different classifiers, my dataset was thoroughly concealed and cleansed for the models training, as i can see in some cases that the external symbols may prove to require difficulties while the analysis via models conjugation. The model is mainly based on the method of obtaining the accuracy via the UNIGRAM, BIGRAM and TRIGRAM structure. Which will be later be described through the article. My work was mainly to determine the two binary classes of positive and negative sentiment structure. I am quite satisfied with my achieved result by going through the related research works of such.