With the development of the Internet, it has increasingly become an indispensable product in people’s lives, but many problems have arisen with it. Internet language violence is one of them. At present, the main Internet users in my country are young people, and online language violence brings extremely serious psychological problems to young students. In order to understand the current mental health of young students and the impact of online language violence on them, this article investigates the students in the city’s no. 1 middle school, filters the data through decision tree analysis, and judges online language through the psychological symptoms self-rating scale. The mental health symptom self-rating scale has the characteristics of large capacity, abundant symptoms, and more accurate description of the subject’s conscious symptoms. It contains a wide range of psychiatric symptoms, from feelings, emotions, thinking, consciousness, and behavior to life habits, interpersonal relationships, eating, and sleeping, and it uses 10 factors to reflect the psychological symptoms in 10 aspects. It has good distinguishing ability for people with psychological symptoms (that is, they may be on the edge of psychological disorder or mental disorder). The chi-square statistical method is used to analyze the basic characteristics of different youth groups of verbal violence. For the research content, the school surveys all schools and adopts questionnaire surveys and case studies to analyze the factors that influence youths’ attribution of online violence behaviors and make recommendations. The impact of violence on young students will be analyzed later through psychological intervention. The results of the study found that online language violence caused serious harm to young people’s psychology and caused students’ depression, anxiety, and other mental illnesses. Active psychological interventions can effectively alleviate students’ mental health. The improvement of students’ mental health is the most obvious. The score improved by nearly 10%.