Table 3.

Correlation performances (average of correlation coefficients evaluated over each sequence) of different audio and lexical features in terms of speaker-dependent (SD) and speaker-independent (SI) sequence, on the development set

WlSC-development	arousal		expectancy		power		valence		average
Baseline Audio	0.097		0.052		0.061		0.085		0.074
	SD	SI	SD	SI	SD	SI	SD	SI	SD	SI
Audio features
TL acoustic	0.340	0.208	0.244	0.235	0.164	0.137	0.340	0.227	0.272	0.202
CL acoustic	0.363	0.216	0.211	0.184	0.170	0.175	0.380	0.203	0.281	0.195
Lexical features
TL PMI	0.190	0.188	0.260	0.202	0.218	0.171	0.205	0.200	0.218	0.190
TL Sparse Lex	0.268	0.208	0.228	0.204	0.146	0.168	0.216	0.302	0.215	0.221