Table 3.

Machine learning classification results of models trained on reading task data with pause features computed using acoustic data or computed using text data.

Reading task	(1) Automatic transcripts AUROCa,b	(2) Manually corrected transcripts AUROCb	(3) Manually corrected transcripts AUROCc	Change in AUROC (3)–(1)
RFd	0.638	0.655	0.662	0.024
GNBe	0.677	0.677	0.693	0.016
LRf	0.589	0.587	0.568	−0.021

^aAUROC: area under the receiver operating characteristic curve.

^bPauses detected from acoustic data.

^cPauses detected from text data.

^dRF: random forest.

^eGNB: Gaussian naive Bayes.

^fLR: logistic regression.