Adaptive multi-view selection for semi-supervised emotion recognition of posts in online student community.pdfVIP

Neurocomputing 144 (2014) 138–150 Contents lists available at ScienceDirect Neurocomputing journal homepage: /locate/neucom Adaptive multi-view selection for semi-supervised emotion recognition of posts in online student community Zongkai Yang a, Zhi Liu a,n, Sanya Liu a,b,nn, Lei Min a, Wenting Meng c a National Engineering Research Center for E-Learning, Central China Normal University, Wuhan 430079, PR China b Key Laboratory of Adolescent Cyberpsychology and Behavior, Ministry of Education, Wuhan 430079, PR China c Wuhan Branch Company, Hikvision Digital Technology Co., Ltd., Wuhan 430074, PR China article info Article history: Received 9 November 2013 Received in revised form 7 March 2014 Accepted 9 May 2014 Communicated by Yongdong Zhang Available online 12 June 2014 Keywords: Semi-supervised emotion recognition Multi-view selection Diversity Feature emotional strength Kernel smoothing Emotion contribution rate abstract In statistical text emotion recognition, semi-supervised learning that can leverage plenty of unlabeled data has drawn much attention in recent years. However, the quality of the training data is typically in?uenced by some mislabeled samples. In this paper, we present a novel co-training method, namely adaptive multi-view selection (AMVS), to improve labeling accuracy of unlabeled samples for semisupervised emotion recognition. In particular, two importance distributions are proposed to construct multiple discriminative feature views. One is the distribution of feature emotional strengths, and the other is the importance distribution of view dimensionality. On the basis of these two distributions, several feature views are iteratively selected from the original feature space in a cascaded way, and corresponding base classi?ers are trained on these views to build a dynamic and robust ensemble. The experimental results on the real-life dataset consisting of moods posts demonstrate the proposed AMVS outperforms conventional multi-view semi-supervi