An approach and software system for integral analysis of destructive paralinguistic phenomena in colloquial speech
Keywords:
speech analysis, speech technologies, computational paralinguistics, destructive phenomena, automatic deception detection in speech, automatic aggression detection in speech, automatic depression detection in speechAbstract
Introduction: There has been a growth in the number of studies devoted to the developing systems for paralinguistic analysis of various destructive phenomena in speech. However, only a few studies consider several phenomena altogether, which is undoubtedly an omission because there are also some theoretical and practical studies describe the relationships between different paralinguistic phenomena. Purpose: to develop an approach and software system for integral analysis of destructive paralinguistic phenomena (deception, aggression, depression) in colloquial speech. Results: we present an approach for integral evaluation of destructive paralinguistic phenomena in speech that takes into account classification results from each of the three methods for detection the abovementioned paralinguistic phenomena and use it to calculate the integral score according to the set of rules. The proposed architecture of software system includes a complex of methods that consists of three methods for detection the abovementioned paralinguistic phenomena and the proposed approach. On the basis of the experimental classification results on detection of destructive paralinguistic phenomena to evaluate the proposed system we can calculate the integral mean F1-measure and integral unweighted average recall which result in 76.8% and 75.0% accordingly. Practical relevance: the use of the proposed architecture of the software system makes it possible to obtain an integral evaluation of destructive paralinguistic phenomena in colloquial speech. Such software system can be applied to the initial assessment of patients in consultation with medical professionals as one of the methods of psychological evaluation as well as other classical approaches for initial assessment of patients (questionaries, tests, etc.).