Fig. 3From: Classification-based spoken text selection for LVCSR language modelingPerplexities with accumulation of groups for SVM-based scoringBack to article page