Table 7.

Performance of classification models using a combination of lexical and different types of non-lexical features according to standard metrics for the task of annotating caregiver interview session transcripts.

Cls.	Model	Acc.	Prec.	Rec.	F1	Kappa
16	CRF	0.654	0.652	0.654	0.653	0.603
	SVM	0.664	0.653	0.664	0.639	0.606
	SVM-PL	0.670	0.658	0.670	0.651	0.614
	SVM-LIWC	0.730	0.730	0.730	0.717	0.686
	SVM-AF	0.738	0.733	0.738	0.727	0.696
19	CRF	0.539	0.541	0.539	0.540	0.492
	SVM	0.545	0.547	0.545	0.535	0.497
	SVM-PL	0.566	0.570	0.566	0.559	0.522
	SVM-LIWC	0.620	0.625	0.620	0.613	0.581
	SVM-AF	0.638	0.639	0.638	0.631	0.601
58	CRF	0.438	0.409	0.438	0.423	0.385
	SVM	0.451	0.420	0.451	0.418	0.414
	SVM-PL	0.480	0.462	0.480	0.456	0.446
	SVM-LIWC	0.459	0.445	0.459	0.429	0.422
	SVM-AF	0.488	0.466	0.488	0.462	0.454