Affective and Contextual Embedding for Sarcasm Detection

• DOI: 10.18653/v1/2020.coling-main.20
• Publication Date: 2021-01-08T13:58:31Z
• Authors: Nastaran Babanejad; Heidar Davoudi; Aijun An; Manos Papagelis
• Keywords: Sarcasm; Word2vec; Autoencoder; Sentiment Analysis

Automatic sarcasm detection from text is an important classification task that can help identify the actual sentiment in user-generated data, such as reviews or tweets. Despite its usefulness, remains a challenging task, due to lack of any vocal intonation facial gestures textual data. To date, most approaches addressing problem have relied on hand-crafted affect features, pre-trained models non-contextual word embeddings, Word2vec. However, these inherit limitations render them inadequate for detection. In this paper, we propose two novel deep neural network detection, namely ACE 1 and 2. Given input passage, predict whether it sarcastic (or not). Our extend architecture BERT by incorporating both affective contextual features. best our knowledge, first attempt directly alter BERT’s train scratch build classifier. Extensive experiments different datasets demonstrate proposed outperform state-of-the-art with significant margins.