Crisis-intervention chatbot for cyberbullying victims

Yu, Dong Yang (2025) Crisis-intervention chatbot for cyberbullying victims. Final Year Project, UTAR.

PDF Download (6Mb)

Abstract

The common manifestation of cyberbullying in hear times has become an acute social problem, with dire impact on emotional health and psychological welfare of its victims, in especially teen and youth groups. Although support networks and counseling services are critical over the long term as a preventive mechanism in their mitigation of long-term impacts, they are not always capable of undertaking prompt, personalized intervention to the victims when their need becomes immediate and personal. This constraint leaves a gap in service provision in instances where members of the population who are in emotional distress can fail to get help promptly. To overcome this obstacle, this project proposes the use of a web-based crisis-intervention chatbot that is specifically dedicated to victims of cyberbullying. The proposed system is a hybrid conversational engine of ChatterBot and a large language model that can generate responses informed by context and an empathetic perspective. The backend is built on Flask, which allows lightweight and efficient deployments with MySQL as database to handle the user session and store conversation history safely. Another distinctive aspect of the system is its sentiment analysis based on Textblob that can recognize the emotions of users with sentiment analysis and inclusion of all typical NLP preprocessing techniques like tokenization, lemmatization, and removal of stopwords. Such processes improve the capabilities of a chatbot to find the intentions of users, present emotional sentiment patterns, and adequately respond to every distinct scenario. As with all user interactions, everything takes place on anonymity and is carefully organized and secure such as preserving privacy. The chatbot can deliver real-time emotional support, coping mechanisms, and references to specific mental health resources, utilizing a combination of natural language processing, sentiment analysis, and large language model intelligence.

Actions (login required)