Attentive User-Engaged Adversarial Neural Network for Community Question Answering

• DOI: 10.1609/aaai.v34i05.6472
• Publication Date: 2020-06-18T08:17:20Z
• Authors: Yuexiang Xie; Ying Shen; Yaliang Li; Min Yang; Kai Lei
• Keywords: SemEval; Similarity (geometry)

We study the community question answering (CQA) problem that emerges with advent of numerous forums in recent past. The task finding appropriate answers to questions from informative but noisy crowdsourced is important yet challenging practice. present an Attentive User-engaged Adversarial Neural Network (AUANN), which interactively learns context information and answers, enhances user engagement CQA task. A novel attentive mechanism incorporated model semantic internal external relations among questions, contexts. To handle noise issue caused by introducing context, we design a two-step denoise mechanism, including coarse-grained selection process similarity measurement, fine-grained applying adversarial training module. evaluate proposed method on large-scale real-world datasets SemEval-2016 SemEval-2017. Experimental results verify benefits incorporating information, show our significantly outperforms state-of-the-art methods.