RTGBO based Recurrent Neural Network for Sarcasm Recognition using Twitter Data

Increases in popularity of social media stages have been meteoric in fresh years. Through social media, a massive quantity of information is being released to the public. This massive amount of freely accessible data has several potential uses. This work aims to discourse issues with the social media dataset, such as its short text nature (with just 140 to 160 characters of text data), its continuous cyclosis nature, its use of abbreviations and slang, and the growing prevalence of sarcasm in online communications. Sentiment analysis and sarcasm recognition are two of the most important jobs in NLP. To communicate sarcasm, one must provide an outside element that reverses the positive and negative emotions involved. There is a risk that data mining efforts may be misclassified if they are fed sarcastic tweets. In this study, we provide a Recurrent Neural Network-based technique for analysing the tone of tweets (RNNs). The proposed approach combines a memory cell-level and network-level localization system for identifying sarcastic and non-sarcastic tweets with an attention mechanism designed to improve the network's performance. It uses an algorithm called Ring-Toss-Game-Based Optimization (RTGBO) to fine-tune the hyper-parameters. The primary focus of RTGBO is to provide a simulation of player behaviour and rules for the ring toss game. One distinguishing feature of the RTGBO procedure is that it has no tuning knobs. Since newspaper headlines often include sarcasm in an effort to draw in readers, the newspaper sarcasm dataset is used for this research. Existing methods are used to verify the suggested classifier in terms of a number of metrics. The model works quite well and accurately identifies which sentences are sarcastic and which are not.