Detecting Suicidal Thoughts in Social Media: A Proposed NLP and Chatbot Framework Informed by a Preliminary Survey

Abstract

This conceptual paper presents a proposed natural language processing (NLP) and chatbot framework for recognizing potentially concerning language in social media posts and directing users toward human support. A preliminary survey provides context on respondents’ platform use, encounters with posts expressing depressed thoughts, responses to such posts, and awareness of support resources. Demographic and platform-use summaries cover seven respondents, while three response items contain five tabulated answers each. Four of seven respondents were aged 25-34; all seven reported using Facebook, and five reported using Instagram. Of the five tabulated answers to the resource-awareness question, four indicated no awareness of support resources or helplines. The proposed workflow links authorized text collection, preprocessing, contextual classification, and a conversational interface that offers support information and routes users toward appropriate services. The survey informs initial design priorities, particularly accessible resource information and a clear route to human help. The contribution is a conceptual framework and a descriptive account of the preliminary survey; subsequent implementation research should examine classification performance, response quality, usability, privacy, and referral outcomes.