TY - JOUR PY - 2024// TI - Exploring the role of first-person singular pronouns in detecting suicidal ideation: a machine learning analysis of clinical transcripts JO - Behavioral sciences (Basel, Switzerland) A1 - Huang, Rong A1 - Yi, Siqi A1 - Chen, Jie A1 - Chan, Kit Ying A1 - Chan, Joey Wing Yan A1 - Chan, Ngan Yin A1 - Li, Shirley Xin A1 - Wing, Yun Kwok A1 - Li, Tim Man Ho SP - e225 EP - e225 VL - 14 IS - 3 N2 - Linguistic features, particularly the use of first-person singular pronouns (FPSPs), have been identified as potential indicators of suicidal ideation. Machine learning (ML) and natural language processing (NLP) have shown potential in suicide detection, but their clinical applicability remains underexplored. This study aimed to identify linguistic features associated with suicidal ideation and develop ML models for detection. NLP techniques were applied to clinical interview transcripts (n = 319) to extract relevant features, including four cases of FPSP (subjective, objective, dative, and possessive cases) and first-person plural pronouns (FPPPs). Logistic regression analyses were conducted for each linguistic feature, controlling for age, gender, and depression. Gradient boosting, support vector machine, random forest, decision tree, and logistic regression were trained and evaluated.

RESULTS indicated that all four cases of FPSPs were associated with depression (p < 0.05) but only the use of objective FPSPs was significantly associated with suicidal ideation (p = 0.02). Logistic regression and support vector machine models successfully detected suicidal ideation, achieving an area under the curve (AUC) of 0.57 (p < 0.05). In conclusion, FPSPs identified during clinical interviews might be a promising indicator of suicidal ideation in Chinese patients. ML algorithms might have the potential to aid clinicians in improving the detection of suicidal ideation in clinical settings.

Language: en

LA - en SN - 2076-328X UR - http://dx.doi.org/10.3390/bs14030225 ID - ref1 ER -