Get started
Work

VoiceAgriNet: Dual-Agent Voice Interaction for Trustworthy Data Collection and Retrieval from Smallholder Farmers

Year: 2026
Type: conference-paper
Abstract: Collecting structured agricultural data from small- holder farmers remains difficult in low-literacy, low-connectivity settings, especially when the interface is voice-based and response time matters. We present VoiceAgriNet, an agricultural data- collection system that adapts a dual-agent RAG architecture for spoken interaction. The background Slow Thinker pre-fetches likely context into a semantic cache. The foreground Fast Talker answers from that cache when a relevant match is available. In our evaluation, VoiceAgriNet improves retrieval speed over a same-hardware FAISS baseline while maintaining safe responses, grounded answers, and accurate structured extraction. The retrieval gain applies only to the retrieval stage; end-to-end latency remains dominated by LLM generation and is reported separately. These results show that proactive semantic caching can make agricultural voice systems more responsive without weakening grounding or safety. Accepted manuscript (post-peer-review, author-generated version). © 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The published version is available at https://doi.org/10.1109/COMPSAC69091.2026.00420 Funding: European Union Horizon Europe programme, Marie Skłodowska-Curie grant agreement No 101073381 (EnTrust — Next-generation of trustworthy agri-data management).
DOI: 10.1109/compsac69091.2026.00420
Language: English
Subfield: Plant Science
Open Access status: green
OpenAlex ID: W7203832372
FWCI:
Cites: 16
Cited by:
Related to: 10