Semi-supervised learning based Chinese dialect identification
Mingliang Gu, Yuguo Xia, Yiming Yang · 2008
The main contribution of this paper is to present the novel information fusing framework. The shifted delta cepstra (SDC) feature and prosodic feature are firstly used to train two base classifiers respectively. Then co-training algorithm presented in semi-supervised learning is employed to improve Chinese dialect identification accuracy. Four Chinese dialects is tested with this system. The experimental results showed that the proposed system outperformed the original GMM based system.