As social media platforms gain traction, it becomes important to monitor the content that is being exchanged. Maintaining clean and engaging medium for interactions becomes a challenge with increased traffic; this is where content policing comes into the picture. Sentiment analysis is one of the strategies that can be employed in order to constantly monitor the content. While a standard classification model can handle straight forward cases, it can become challenging for them to handle when an ambiguous case is occurred. Such limitations of traditional models pose the need for better techniques in order to handle the ambiguities. The aim of this work is to handle such uncertain cases with the help of more lenient machine learning models, which provides better trade-off between maximizing the margin and minimizing the penalty. In this work encoding methods are used to perform tokenization and variants of fuzzy based models, soft-margin SVM, LSTM and RBF Network are used for classification. The results show that the combination of WordPiece encoding along with LSTM are outperforming when compared to other soft-computing models.