Scholarlink

The adoption of Large Language Models in Indian agriculture presents an unprecedented opportunity to enhance farmer education, decision-making, and access to vital agricultural knowledge. This paper introduces AgriLLM-India, a domain-specific chatbot tailored to the Indian context, incorporating local languages, region-specific crops, and government policy frameworks. Utilizing a Retrieval-Augmented Generation architecture, AgriLLM-India integrates a multilingual dataset derived from Indian agricultural research, government schemes (like PM-KISAN, eNAM), and agronomic best practices. The framework leverages Facebook AI Similarity Search for semantic search and compares the performance of three LLMs—ChatGPT-4o Mini, Gemini 1.5 Flash, and Mistral-7B-Instruct-v0.2—on their ability to respond accurately and contextually to queries in Hindi, Telugu, and English. A case study on cotton cultivation in Telangana highlights the chatbot’s ability to provide region-specific insights on irrigation, pest control, and subsidy access. Results demonstrate that ChatGPT-4o Mini with RAG achieved the highest accuracy and contextual relevance, albeit with slightly higher response latency. The findings underscore the value of LLMs tailored to Indian agricultural needs and call for scalable AI models that incorporate real-time data and multilingual support to democratize agricultural advisory services.

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

AgriLLM-India: Empowering Indian Farmers Through Domain-Specific Large Language Models for Agricultural Knowledge and Decision Support

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Vijay Kumar Damera,
  • Deepthi Kalavala

摘要

The adoption of Large Language Models in Indian agriculture presents an unprecedented opportunity to enhance farmer education, decision-making, and access to vital agricultural knowledge. This paper introduces AgriLLM-India, a domain-specific chatbot tailored to the Indian context, incorporating local languages, region-specific crops, and government policy frameworks. Utilizing a Retrieval-Augmented Generation architecture, AgriLLM-India integrates a multilingual dataset derived from Indian agricultural research, government schemes (like PM-KISAN, eNAM), and agronomic best practices. The framework leverages Facebook AI Similarity Search for semantic search and compares the performance of three LLMs—ChatGPT-4o Mini, Gemini 1.5 Flash, and Mistral-7B-Instruct-v0.2—on their ability to respond accurately and contextually to queries in Hindi, Telugu, and English. A case study on cotton cultivation in Telangana highlights the chatbot’s ability to provide region-specific insights on irrigation, pest control, and subsidy access. Results demonstrate that ChatGPT-4o Mini with RAG achieved the highest accuracy and contextual relevance, albeit with slightly higher response latency. The findings underscore the value of LLMs tailored to Indian agricultural needs and call for scalable AI models that incorporate real-time data and multilingual support to democratize agricultural advisory services.