Correlating natural language processing and automated speech analysis with clinician assessment to quantify speech-language changes in mild cognitive impairment and Alzheimer’s dementia

Table 4 Incoherence (clinician-rated) and correlations (p<0.05) with speech variables, ranked by correlation coefficient (Spearman’s |ρ|)

Variable type	Variable description	ρ	F1	F2
Syntactic	Use of past tense verb phrases with noun phrases	0.51	0.69
Lexical	Average age of acquisition score of all words	0.47		− 0.64
Acoustic	Speech rate	− 0.47
Acoustic	Average word duration	0.47	0.64
Lexical	Average age of acquisition score of nouns	0.46		− 0.65
Lexical	Average valence score of nouns	− 0.43		0.79
Acoustic	Skewness of the second derivative of the 4th MFCC	− 0.42	− 0.87
Semantic	Proportion of subject words used	− 0.42
Syntactic	Use of 3rd person singular present verb phrases with prepositional phrases	− 0.41
Acoustic	Kurtosis of the second derivative of the 4th MFCC	0.41	0.98
Syntactic	Use of prepositional phrases with noun phrases	− 0.40
Acoustic	Skewness of the second derivative of the log energy	− 0.39	− 0.96
Lexical	Use of comparative adjectives	0.38
Acoustic	Kurtosis of the second derivative of the log energy	0.38	0.95
Semantic	Semantic similarity of description to picture content (max cosine distance, 300-dim word vectors)	0.38
Semantic	Average valence score of all words	− 0.37		0.82
Syntactic	Use of singular present verb phrases with prepositional phrases	− 0.37
Syntactic	Use of verb phrases with noun phrases	0.37
Semantic	Proportion of subjects in picture described	− 0.37
Lexical	Average arousal score of nouns	− 0.36

Columns F1 and F2 indicate which variables were assigned to each factor, and the factor loading scores. Variables that do not have values in columns F1 and F2 are correlated with incoherence but were not included in a factor based on the exploratory factor analysis

ISSN: 1758-9193