Sarcasm Detection Using Deep Convolutional Neural Networks: A Modular Deep Learning Framework

TL;DR

The paper tackles sarcasm detection in text, a task challenged by reliance on context and tone. It proposes a modular deep learning framework that fuses four detectors—Sentiment, Contextual Embedding, Linguistic Features, and Emotion Detection—via DCNNs and BERT to form a unified representation for sarcasm classification. A conceptual multimodal case study demonstrates that integrating visual cues with text (BERT+DenseNet) improves accuracy over text alone, illustrating the practical potential for applications in moderation and conversational agents. The approach emphasizes extensibility, explainability, and fairness, outlining a path toward multilingual and audio-visual enhancements in future work.

Abstract

Sarcasm is a nuanced and often misinterpreted form of communication, especially in text, where tone and body language are absent. This paper proposes a modular deep learning framework for sarcasm detection, leveraging Deep Convolutional Neural Networks (DCNNs) and contextual models such as BERT to analyze linguistic, emotional, and contextual cues. The system integrates sentiment analysis, contextual embeddings, linguistic feature extraction, and emotion detection through a multi-layer architecture. While the model is in the conceptual stage, it demonstrates feasibility for real-world applications such as chatbots and social media analysis.

Figures (5)

• Figure 1: Illustration of a Basic Neural Network
• Figure 2: Architecture of a Convolutional Neural Network
• Figure 3: Structure of a Deep Convolutional Neural Network
• Figure 4: Proposed Modular System Architecture for Sarcasm Detection
• Figure 5: Model Accuracy Comparison on Twitter Dataset