Stance-level Sarcasm Detection with BERT and Stance-centered Graph Attention Networks

Computational Linguistics (CL) associated with the Internet of Multimedia Things (IoMT)-enabled multimedia computing applications brings several research challenges, such as real-time speech understanding, deep fake video detection, emotion recognition, home automation, and so on. Due to emergence machine translation, CL solutions have increased tremendously for different natural language processing (NLP) applications. Nowadays, NLP-enabled IoMT is essential its success. Sarcasm a recently emerging artificial intelligence (AI) NLP task, aims at discovering sarcastic, ironic, metaphoric information implied in texts that are generated IoMT. It has drawn much attention from AI community. The advance sarcasm detection techniques will provide cost-effective, intelligent way work together devices high-level human-to-device interactions. However, existing approaches neglect hidden stance behind texts, thus insufficient exploit full potential task. Indeed, stance, i.e., whether author text favor of, against, or neutral toward proposition target talked text, largely determines text’s actual orientation. To fill gap, this research, we propose new task: stance-level (SLSD), where goal uncover author’s latent based on it identify polarity expressed text. We then an integral framework, which consists Bidirectional Encoder Representations Transformers (BERT) novel stance-centered graph networks (SCGAT). Specifically, BERT used capture sentence representation, SCGAT designed specific target. Extensive experiments conducted Chinese sentiment dataset created SemEval-2018 Task 3 English dataset. experimental results prove effectiveness framework over state-of-the-art baselines by large margin.