Scholarlink

Filter bubbles and echo chambers on social networking services (SNS) often limit users’ exposure to diverse perspectives. To address this issue, we propose a method for extracting multiple viewpoints in real time during SNS searches and presenting them on an interactive positioning map displayed alongside the list of posts. This map visually organizes posts along axes that represent diverse perspectives, which are generated by a large language model (LLM) based on search keywords and retrieved content. The system, developed using the X (formerly Twitter) API and GPT-4o API, was evaluated in a user study involving 10 university students and 17 search keywords, each retrieving 100 posts. The study revealed a significant change in user behavior: participants explored posts primarily through interactions with the map rather than by scrolling through traditional lists.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

SNS Search System that Uses a Large Language Model to Display Posts on an Interactive Map from Various Perspectives

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Tomonari Kamba

摘要

Filter bubbles and echo chambers on social networking services (SNS) often limit users’ exposure to diverse perspectives. To address this issue, we propose a method for extracting multiple viewpoints in real time during SNS searches and presenting them on an interactive positioning map displayed alongside the list of posts. This map visually organizes posts along axes that represent diverse perspectives, which are generated by a large language model (LLM) based on search keywords and retrieved content. The system, developed using the X (formerly Twitter) API and GPT-4o API, was evaluated in a user study involving 10 university students and 17 search keywords, each retrieving 100 posts. The study revealed a significant change in user behavior: participants explored posts primarily through interactions with the map rather than by scrolling through traditional lists.