Abstract:A large number of individuals are suffering from suicidal ideation in the world. There are a number of causes behind why an individual might suffer from suicidal ideation. As the most popular platform for self-expression, emotion release, and personal interaction, individuals may exhibit a number of symptoms of suicidal ideation on social media. Nevertheless, challenges from both data and knowledge aspects remain as obstacles, constraining the social media-based detection performance. Data implicitness and sparsity make it difficult to discover the inner true intentions of individuals based on their posts. Inspired by psychological studies, we build and unify a high-level suicide-oriented knowledge graph with deep neural networks for suicidal ideation detection on social media. We further design a two-layered attention mechanism to explicitly reason and establish key risk factors to individual's suicidal ideation. The performance study on microblog and Reddit shows that: 1) with the constructed personal knowledge graph, the social media-based suicidal ideation detection can achieve over 93% accuracy; and 2) among the six categories of personal factors, post, personality, and experience are the top-3 key indicators. Under these categories, posted text, stress level, stress duration, posted image, and ruminant thinking contribute to one's suicidal ideation detection.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to detect suicidal ideation on social media. Although social media provides a large amount of data on personal expression and interaction, the implicitness and sparseness of these data make it very difficult to directly detect suicidal ideation from social media content. Specifically:
1. **Data implicitness**: Due to the characteristics of social media, users' language and visual expressions are often implicit, reserved, or even contrary to reality. For example, in public normal posts, users may show happy emotions, but in some hidden "tree holes" (referring to the comment section of a deceased user, where other users with suicidal tendencies will share their true inner feelings), they may show severe despair and suicidal thoughts.
2. **Data sparseness**: Some people may be unwilling to actively express themselves on social media due to habits, personalities, or emotions, especially when they feel desperate, lonely, or have suicidal thoughts. In addition, people with suicidal tendencies tend to delete early posts related to suicide to hide their true inner intentions.
To address these problems, the paper proposes the following solutions:
- **Constructing a personal knowledge graph**: By integrating the results of psychological research, a high - level, suicide - oriented knowledge graph is constructed, and combined with a deep neural network to detect suicidal ideation on social media. This knowledge graph contains data on multiple aspects such as personal information, personality, experiences, posting behaviors, emotional expressions, and social interactions.
- **Designing a two - layer attention mechanism**: The attribute attention and neighbor attention mechanisms are introduced to explicitly reason and determine the key risk factors for individual suicidal ideation. This mechanism can adaptively adjust the contribution weights of different risk factors (such as attributes and neighbor users) to individual suicidal ideation.
- **Experimental verification**: Experiments were carried out on Weibo and Reddit platforms. The results show that by using the constructed personal knowledge graph and the two - layer attention mechanism, the accuracy and F1 - value of suicidal ideation detection both exceed 93%. Among them, posts, personalities, and experiences are the top three key indicators affecting suicidal ideation detection, while specific texts, stress levels, stress durations, posted pictures, and rumination are the most important contributing factors.
In general, through constructing a personal knowledge graph and designing a two - layer attention mechanism, this paper effectively addresses the challenges brought by the implicitness and sparseness of social media data and significantly improves the performance of suicidal ideation detection.