Artificial intelligence-driven Vietnamese-language chatbot for pest diagnosis and pesticide guidance in crop production

Abstract

Digital advisory systems have emerged as an important solution for improving crop management in regions where farmers have limited access to timely and reliable technical support. This study developed and evaluated an artificial intelligence-driven chatbot that uses Vietnamese natural language processing and a structured plant protection knowledge base to provide guidance on pest diagnosis, nutrient management, and pesticide use. System requirements were identified through interviews with 30 farmers and 10 agricultural students. A comprehensive knowledge base was constructed from national plant protection guidelines and international technical sources and validated by specialists from provincial plant protection agencies. The chatbot was built on a modular architecture that included an NLP component, a dialogue management module, and image-supported advisory templates, and was deployed on Facebook Messenger. Technical testing over 72 hours showed stable system performance, with an average response time of 1.3 seconds, an intent-recognition accuracy of 85 percent, and an automation rate of 80 percent. Field testing with 30 farmers generated 431 valid queries and demonstrated strong user engagement, particularly for pest diagnosis and pesticide preparation. User evaluations showed high clarity (4.4), usability (4.3), usefulness (4.2), and overall satisfaction (4.3). Qualitative feedback indicated that farmers valued the standardized, stepwise guidance and the absence of commercial pesticide promotion. The system’s ability to deliver consistent and scientifically validated information suggests its potential as a practical digital extension tool for Vietnamese agriculture. The findings highlight opportunities to integrate more advanced NLP models, region-specific data, and image-based diagnostic features to further enhance advisory accuracy and usability.