Dissertations, Theses, and Capstone Projects

Date of Degree

6-2025

Document Type

Thesis

Degree Name

M.A.

Program

Linguistics

Advisor

Rivka Levitan

Committee Members

Sarah Ita Levitan

Jason Kandybowicz

Keywords

NLP, depression detection, machine learning, italian language

Abstract

Depression does not always speak in obvious ways. It often hides in silence, in brief pauses, and in the quiet texture of language. While the prosodic and linguistic characteristics of depression have been extensively studied in English and other languages, they remain underexplored in Italian, with this being a first focused analysis. This study investigates how depression manifests in the speech of Italian speakers, using transcriptions from a clinical corpus. Drawing on emotional, structural, and cognitive-linguistic features, including first person verb use, sentence complexity, verb tense, and negation, the study evaluates how these markers differ between individuals with and without depression. Even small changes, such as favoring the past tense over the present, appeared to carry psychological meaning. A baseline model trained on these features showed promising results, indicating that minor linguistic variations can reflect deeper emotional states. The study also trained a deep learning model known as BERTino, which achieved an accuracy rate of 92%. Unlike traditional models, this transformer was able to pick up on more subtle expressions of emotional distress, signals that may escape both listener and speaker. This study sheds light on how depression surfaces in the language of Italian speakers and highlights the potential of combining psychologically grounded features with deep learning to better understand what is often left unsaid.

Download

Share

COinS