查看文章

Over the last decades, a growing body of linguistic studies have been devoted to the clinical domain (Perkins, 2011), while the amount of experimental linguistic research focusing on neuroscience and mental health has increased exponentially during the last few years. Considering that many of the factors underlying cognitive and neuropsychiatric disorders may yield to late symptoms that are hard to foresee, it is often difficult to predict the existence of a presence or risk of a disease, as well as the disease’s trajectory. In this context, interdisciplinary approaches gain increasing popularity, and the analysis of complex behavior—such as speech and language—emerges as a natural candidate to identify and analyse the extent to which a given neuropathology can impact the cognitive system at the very early stages. In this context, the development of cognitive evaluation and intervention tools focusing on linguistic biomarkers becomes a critical scientific arena both in and outside the clinic and laboratory (see Petrizzo and Popolo, 2020).

Recent international research has demonstrated that automated collected and analyzed quantitative linguistic features, easily extractable from a patient’s verbal productions, can be very useful in separating people with various cognitive or mental impairment from healthy subjects, even at a very early stage (see Bedi et al., 2015), and even to predict the outcomes of clinical interventions (see Carrillo et al., 2018). In this line, machine learning-based language technology methods and tools based on artificial intelligence are particularly promising to address this task (Locke et al., 2021; Sigman et al., 2021). Indeed …