Detecting sarcasm in multi-domain datasets using convolutional neural networks and long short term memory network...

Area of Science:

• Natural Language Processing
• Computational Linguistics
• Artificial Intelligence

Background:

• Sarcasm detection is challenging due to the use of positive language for negative sentiment expression.
• Existing studies often rely on limited, single-domain datasets.
• A multi-domain approach is needed for robust sarcasm detection.

Purpose of the Study:

• To develop and evaluate a hybrid model for sarcasm detection.
• To address the limitations of single-domain datasets by using a combined Twitter and News Headlines dataset.
• To compare the proposed model's performance against traditional machine learning algorithms and state-of-the-art approaches.

Main Methods:

• A hybrid approach utilizing Convolutional Neural Networks (CNN) for feature extraction and Long Short-Term Memory (LSTM) for classification.
• Feature representation using Term Frequency-Inverse Document Frequency (TF-IDF), Bag of Words (BoW), and Global Vectors for Word Representations (GloVe).
• Performance evaluation using metrics like accuracy, precision, recall, and F1-score, compared against Random Forest, Support Vector Classifier, Extra Tree Classifier, and Decision Tree algorithms.

Main Results:

• The proposed hybrid CNN-LSTM model achieved an accuracy of 91.60%.
• The model significantly outperformed traditional machine learning algorithms in sarcasm detection.
• The hybrid model demonstrated superior precision, recall, and F1 scores compared to state-of-the-art methods.

Conclusions:

• The developed hybrid CNN-LSTM model is accurate and robust for sarcasm detection.
• The model's effectiveness is validated on a challenging multi-domain dataset.
• This approach advances the capabilities of automated sarcasm detection in diverse online text.