A Method for the Identification of Multiple Labels in Social Media Text using CNN Approach

People are able to communicate their thoughts and perspectives with one another through various online settings, including social networks, news websites, and blogs. Because so many individuals instantaneously post their thoughts on the internet, the number of submissions that have been submitted has increased dramatically. It is critical for establishments and organizations to do an effective and speedy analysis of the massive amounts of data at their disposal in order to provide summaries that discuss the perspectives or sentiments of individuals. In this article, we offer a scalable framework that makes sentiment classification by assessing the compound probability scores of the most generally used methods in sentiment analysis using a fuzzy inference mechanism in a CNN-style way. This evaluation is carried out in order to make the framework applicable to a broad range of situations. The valance of an individual's attitude may be estimated by the designed fuzzy inference system by calculating the compound scores of the valance aware dictionary, word embedding, and count vectorization procedures. The suggested technique differs from the convolutional neural networks (CNNs) in that it permits weighting of base learners and combines the strengths of each algorithm using fuzzy rules. Another distinction is that the proposed method is more general. The process of estimating sentiment based on textual data can be handled as either a two-class (positive and negative) or a three-class (positive, neutral, and negative) problem. The former has less challenges than the latter. We carried out the experimental study on four different tagged social network data sets, and discovered that the suggested strategy delivers gains in accuracy for both 2-class and 3-class classifications.