Detecting depression in speech using verbal behavior analysis: a cross-cultural study

Language analysis has proven to be a reliable methodology for discriminating depressed people from healthy subjects; the present study investigates differences in the linguistic content of spoken interactions from depressed and healthy subjects belonging to three different European areas: Northern Ireland, Italy, and Russia.

The speech of 241 participants (65 native English speakers, 108 native Italian speakers, and 68 native Russian speakers) was analyzed, using the computerized text analysis tool LIWC (Linguistic Inquiry Word Count).

In line with the current literature, it was observed that depressed subjects tended to use more first-person singular pronouns, speak less, use a more negative tone while speaking and use more words related to negative emotions and anxiety compared to healthy controls. Our study also highlighted some innovative findings, such as depressed subjects' greater spontaneity and tendency to speak with less self-censorship compared to healthy participants, as well as a tendency to adopt a type of thinking defined as "informal" rather than analytic. Moreover, our study is the first, at least to the best of our knowledge, comparing speech content of depressed participants belonging to three European areas: Western Europe (Northern Ireland), Southern Europe (Italy) and Eastern Europe (Russia).

Data collected through the present study could be useful in providing guidelines for the design of autonomous systems able to detect early signs of mood changes and depression through the analysis of interactional exchanges. The final aim is to provide automated and cost-effective technological interventions to be used in health care centers, as well as by mental health professionals, such as psychologists, psychiatrists, psychotherapists, therefore with the aim to provide assistance, jointly with the clinician's expertise, in the process of diagnosing depression.