Project Overview
Voice-to-Action System for Agricultural Data Entry
About This Project
This graduation project focuses on developing a voice-to-action system for AgroVision's Sows application, enabling farmers to record livestock data through speech commands rather than manual touchscreen input.
Student Information
- Name: Sebastian van der Westhuizen
- Student Number: 505951
- Program: HBO-ICT, Software Engineering
- Duration: January 2026 - May 2026 (20 weeks)
- Company: AgroVision B.V.
- Company Supervisor: Kristian Koch
Problem Statement
Farmers cannot efficiently record data while working hands-on with animals, as the Sows application requires manual touchscreen input. This project aims to solve this by implementing a voice-based data entry system.
System Components
The voice-to-action system consists of three main components:
1. Speech Recognition
Converting voice commands to text accurately, supporting multiple languages (initially English, with future expansion to German, Dutch, and Spanish).
2. AI Interpretation
Using AI/NLP to understand intent and extract entities from transcribed text. For example, interpreting "record the weight of 7 piglets at 25 kilos" into:
- Action: record weight
- Animal: piglets
- Quantity: 7
- Weight: 25 kilos
3. Action Pipeline
Verifying extracted data, handling incomplete commands, and triggering the appropriate functions in the existing Sows application.
Technology Stack
- Platform: .NET MAUI (existing Sows application)
- Cloud Services: Azure (Speech Services, OpenAI)
- Target: Mobile devices (iOS/Android)
Supported Features
The voice-to-action system will support the following Sows application features:
| Feature | Description |
|---|---|
| Sow Card | View/manage individual sow information |
| Entry | Register new animals entering the farm |
| Exit | Register animals leaving the farm |
| Service | Record insemination/breeding services |
| Farrowing | Record birth events and litter data |
| Weaning | Record weaning events |
| Condition | Record body condition scores |
| Treatment | Record medical treatments and medications |
| Relocation | Record animal movements between locations |
Research Questions
Main Question: How can a voice-to-action system be designed and integrated into the Sows .NET MAUI application to enable farmers to record livestock data through speech?
Key Sub-questions:
- What are the functional and non-functional requirements for the system?
- Which speech recognition technologies are suitable for .NET MAUI?
- How can AI/NLP be applied to interpret commands and extract structured data?
- Is offline AI interpretation feasible on-device?
- How should the action pipeline validate data and trigger functions?
- How can the complete system be integrated while maintaining code quality?
About AgroVision
AgroVision B.V. is a leading agricultural software company founded in 1986, serving customers in 40+ countries with a team of 175+ employees. They provide specialized software solutions for pig farmers, dairy farmers, crop management, and financial tools for the agricultural sector.