Sarcasm detection on news headline using multilayer bidirectional-LSTM with glove embeddings

The existence of sarcasm in sentences causes results in the determination that the sentiment of a text is inaccurate because sarcasm is difficult to analyze automatically, even by humans. Sarcasm Detection is a growing research area in Natural Language Processing (NLP) and is implemented for various domains. This research proposed sarcasm detection using Multi-Layer Bidirectional Long Short-Term Memory (BiLSTM). The use of Multi-Layer BiLSTM accurately results in classification text and reduces overfitting. Then, Glove can increase the accuracy of system performance because GloVe constructs an explicit word-context or word co-occurrence matrix using statistics across the whole text corpus. The steps used are text pre-processing, splitting the dataset, embedding, modeling, implementation, and evaluation using a confusion matrix. The result of Multi-Layer BiLSTM with Glove Embeddings to detect sarcasm in a sentence can produce a high accuracy of 95.41% and improve 5.71% from previous research with the same dataset.