ENHANCING AGRICULTURAL DECISION-MAKING WITH COMBINED LANGUAGE MODELS

Agricultural decision-making depends on timely and accurate knowledge, yet many farmers lack access to expert guidance, leading to reduced productivity and crop losses. This paper presents a lightweight and scalable agricultural question-answering system that utilizes DeBERTa-v3 sentence embeddings for semantic representation and FAISS-based similarity search for efficient answer retrieval. Unlike generative approaches, the system retrieves responses from a verified knowledge base, ensuring factual accuracy and eliminating hallucinated outputs. The model is optimized for CPU-only environments, making it suitable for deployment in resource-constrained settings. Experimental results demonstrate improved retrieval precision, faster inference, and higher reliability compared to conventional methods. The proposed system provides an effective AI-driven advisory solution to enhance accessibility and support sustainable agricultural practices.