A Comprehensive Review of Automatic Methods for Suicidal Ideation Detection

Suicide is a complex health concern that affects not only individuals but society as a whole. Its prevention is possible but to be effective it has to consider the medical and technological resources available for its evaluation. Nowadays, computational advancements have enabled new forms of study of this phenomenon. Particularly, the study of suicidal ideation using computer-based techniques primarily involves two main technical approaches: text-based classification and deep learning. Furthermore, it heavily relies on creating and using custom datasets with social media textual data. However, some publications have utilized public information (e.g. records from a particular healthcare provider) in their studies. In this paper, a comprehensive overview of current advancements in automatic suicidal ideation detection, focusing on computer science techniques from 2020 to 2024, is provided. Particularly, it evaluates existing and innovative methodologies, datasets, and limitations in the field to give a proper analysis based on the PRISMA methodology of the current state-of-the-art of this specific task.