Enhanced Sarcasm Detection using Support Scalar Lexical Sematic Analyses with Bi-Directional Gated Recurrent Neural Network for Social Media Content Analysis

Sarcasm identification is a tremendous one for finding hidden knowledge in social media sentence to find the importance of message from different point of view. Traditional techniques failed to analyze the lexical importance of sarcastic terms and feature limits due to high false positive terns of key extraction leads poor accuracy. To resolve this problem, we propose a Enhanced sarcasm detection using Support scalar lexical sematic analyses (SSLSA) with Bi-directional Gated Recurrent Neural Network (BGRNN) for social media content analysis. Initially the preprocessing is carried out key-term normalization to remove the non-determinant words from sentences. Then sarcastic impact rate is extracted from sentence to find the sarcastic key terms. By applying Support Scalar Lexical Sematic Analyses (SSLSA) identify the term ontological feature limits of sarcastic class terms for classification. The selective classes are input progress to Intra class Bi-Directional Gated Recurrent neural Network (Bi-GRNN) which generates the neural network with several iterated neurons. Each neuron is trained with the extracted features where each estimates sentiment weight from text and images to detect the emotion. According to the output obtained from this neural network, the method identifies the emotion based on sarcasm detection in different class as well up to 95.6 % improvement than existing methods. Based on the weight the social media terms be categorized into positive, negative, neutral to recommend productive support.