Suicide Classificaction for News Media Using Convolutional Neural Network (2103.03727v1)

Abstract: Currently, the process of evaluating suicides is highly subjective, which limits the efficacy and accuracy of prevention efforts. AI has emerged as a means of investigating large datasets to identify patterns within "big data" that can determine the factors on suicide outcomes. Here, we use AI tools to extract the topic from (press and social) media text. However, news media articles lack of suicide tags. Using tweets with hashtags related to sucide, we train a neuronal model which identifies if a given text has a suicidade-related contagion. Our results suggest a high level of the impact of mediatic into suicide cases, and a intrinsic thematic relationship of suicide news. These results pave the way to build more interpretable suicide data, which may help to better track, understand its origin, and improve prevention strategies.