Vajratiya Vajrobol
Biography
Vajrobol Vajratiya Public Relations Department of Thailand
Research Interest
Fake News Deep Verify: An LLM-Powered Application for Real-Time Fake News Analysis in Thai and English
Abstract
The rapid proliferation of digital misinformation presents a critical challenge to information integrity. This paper details the design and implementation of a novel, bilingual (Thai/English) fact-checking application engineered to provide rapid, AI-driven credibility analysis of user-submitted content. The system employs a multimodal input architecture, accepting text, speech, PDF documents, and image files (PNG, JPG). At its core, the application leverages the gemini-2.5-flash large language model, integrated with Google Search for real-time information grounding. We introduce a structured prompting methodology that constrains the model's output to a predefined JSON schema, ensuring consistent and parsable results that include a definitive verdict (Credible, Misleading, Fake), a quantitative confidence score, a detailed analysis with inline citations, and identification of relevant authorities. The processed data is rendered into an interactive and accessible web-based dashboard, featuring on-the-fly translation, text-to-speech capabilities, and verifiable source linkage. To facilitate the dissemination of fact-checked information, the system generates shareable artifacts, including a summary PNG infographic and a comprehensive PDF report. This work demonstrates a practical and robust framework for applying grounded generative AI to combat misinformation, offering a scalable tool that empowers users to critically evaluate digital content.
Keywords: Fact-Checking, Misinformation, Generative AI, Google Gemini,Multimodal Analysis, Bilingual
