Irony detection is a specific case of sentiment analysis that aims to recognize a brief of text as ironic or nonironic. The significance of irony detection is to determine incongruity between positive words and the negative context in the sentence. Traditional approaches for irony detection are unable to identify the context incongruity in the text.
Transfer learning in irony detection is developed to address the issue of identifying context incongruity. Transfer learning is an approach of deep learning that enables the model to transfer its knowledge from related tasks. Transfer learning helps identify the explicit and implicit incongruity in the texts by utilizing external sentiment knowledge from associated sentiment resources to train the deep neural network for irony detection. Deep transfer learning for irony detection learns deep features representation of the text and effectively detects the context incongruity using transferred knowledge.
• Precise irony detection is essential for social media analysis and security services.
• Ineffectual irony detection contributes to low performance for sentiment analysis, as it causes polarity reversal.
• In the text classification tasks, irony detection is a demanding task that needs deducing the hidden, ironic intent not attained by syntactic or semantic analysis of the text contents.
• Transfer learning is a productive approach for enhancing irony detection by utilizing sentiment knowledge from external resources.
• Deep learning in irony detection extracts the deep sentimental polarity features from the ironic expressions.
• Deep transfer learning employs a deep learning model to transfer deep sentiment features for detecting incongruity in the irony statements.
• Irony detection using deep transfer learning is the most effective way to detect implicit and explicit context incongruity in irony expression.