Using Machine Learning Algorithms to Detect Suicide Risk Factors on Twitter

Suicide Risk Assessment Using Machine Learning and Social Networks: a Scoping Review

“…From the full-text review, 16 articles were then selected for inclusion [26,[42][43][44][45][46][47][48][49][50][51][52][53][54][55][56]. The flow diagram representing the search process is shown in Fig.…”

“…Fodeh et al 2018 [48] proposed a suicidal ideation detection framework that requires a minimum human effort in annotating data by incorporating unsupervised discovery algorithms. This study includes LSA, LDA, and NMF to identify topics.…”

“…Most of the included studies propose models to classify collected text into suicide-related categories. Text classification is the most common objective in the included studies (12/16, 75%) [26,[42][43][44][45][46][47][48][51][52][53][54]. A score estimation of suicidal probability based on post content was proposed in one of the included studies (1/16, 6,25%) [56].…”

“…Feature extraction and feature selection were identified as main objectives in four different studies (4/16, 25%) [48,50,53,56]. The remaining categories (Entity Recognition, Theme Identification and Emotion Recognition) were identified only in a study (1/16, 6.25%) [47,49,55].…”

“…The remaining categories (Entity Recognition, Theme Identification and Emotion Recognition) were identified only in a study (1/16, 6.25%) [47,49,55]. In total, 4 of 16 studies (25%) can be grouped in two categories, involving text classification (3/4) [47,48,53] or score estimation (1/4) [56].…”