الفهرس | Only 14 pages are availabe for public view |
Abstract Suicide is a serious issue in modern societies worldwide. Suicidal ideation refers to individuals’ intention to end their lives. Suicide can be caused by several risk factors. Long-term exposure to negative feelings, life events, hopelessness, anxiety, depression, and social isolation are the most common risk factors. One of the most effective methods of suicide prevention is the early detection of suicidal ideation. If those risk factors are identified early, suicide attempts can be reduced or prevented. The advancements in online communication and social network sites also provide a platform for people to express their real-world sufferings and feelings, which serves as a source for suicidal ideation detection. User-generated content, particularly user-posted text, reveals valued information about users’ statuses and reflects their mental states. In the fields of psychology and NLP the detection of suicidal ideation, primarily through online social media, has become a significant research topic. This thesis presents a new framework to improve the performance of suicide ideation detection. The proposed framework covers a variety of suicide ideation detection problems in the literature review such as focusing solely on the level of words without regard for context meaning and using plain feature sets that are insufficient for detecting suicidal ideation. It consists of four phases. |