1. High-accuracy and low-latency speech recognition with two-head con-textual layer trajectory LSTM model;li;Proc ICASSP,0
2. Restructuring of deep neural network acoustic models with singular value decomposition
3. Random forests-based con-fidence annotation using novel features from confusion network;xue;Proc ICASSP,0
4. Predicting speech recognition con-fidence using deep learning with word identity and score features;huang;Proc ICASSP,0