Sarcasm detection using deep learning and ensemble learning

Across the globe, there is a noticeable upward trend of incorporating sarcasm in everyday life. This trend can be easily attributed to the frequent use of sarcasm in everyday life, but more specifically to social media and the Internet. This study aims to bridge the gap between human and machine intelligence to recognize and understand sarcastic behavior and patterns. The research is based on using various neural techniques, namely Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Baseline Convolutional Neural Networks (CNN) in an ensemble model to detect sarcasm on the internet. In order to improve the precision of the proposed model, the required dataset is also prepared on different previously trained word-embedding models like fastText, Word2Vec, and GloVe, etc., and their accuracies are compared. The aim is to be able to quantify the overall sentiment of the writer as positive or negative / sarcastic or non-sarcastic to ensure that the correct message is received to the intended audience. The final study revealed that the proposed ensemble model with word embeddings outperformed the other state-of-the-art models and deep learning models considered in this study with an accuracy of around 96% for News Headlines dataset, 73% for Reddit dataset, and amongst our proposed ensemble models, Weighted Average Ensemble gave the highest accuracy of around 99% and 82% for both the datasets respectively. Ensemble model used in our study improvised the stability, precision and predictive power of the proposed model.