jaykeshav/domain-aware-odia-ner-for-smart-agriculture

Domain-Aware Named Entity Recognition for Smart Agricultural Query Processing in Low-Resource Odia Language

Overview

This project presents a Domain-Aware Named Entity Recognition (NER) system for smart agricultural query processing in the low-resource Odia language.

The system uses a fine-tuned XLM-RoBERTa model to identify agricultural entities from Odia farmer queries and maps the extracted entities to a domain-specific agricultural knowledge base to generate relevant recommendations.

Project Links

Live Demo

Try the Application

Hugging Face Model

View Model

BIO Tagged Dataset

View Dataset

Agricultural Knowledge Base

View Dataset

Features

• —Odia Language Agricultural Query Processing
• —Custom BIO-tagged Agricultural Dataset
• —Fine-tuned XLM-RoBERTa Model
• —Entity Extraction:
• —Crop
• —Disease
• —Symptom
• —Pest
• —Knowledge Base Integration
• —Agricultural Recommendation Generation
• —Interactive Gradio Demo

System Workflow

Farmer Query (Odia) ↓ Named Entity Recognition (XLM-RoBERTa) ↓ Entity Extraction ↓ Knowledge Base Matching ↓ Agricultural Recommendation

Model Information

Model: XLM-RoBERTa

Task: Named Entity Recognition (NER)

Entity Labels:

• —CROP
• —DISEASE
• —SYMPTOM
• —PEST

Performance

Knowledge Base

The extracted entities are matched against a custom agricultural knowledge base containing:

• —Crop Information
• —Diseases
• —Pests
• —Symptoms
• —Recommended Medicines
• —Dosage Information
• —Preventive Measures
• —Agricultural Solutions

Team Members

• —Chittaranjan Panda
• —Jay Keshav
• —Pritish Kumar Nanda
• —Debansh Panda

Academic Information

Final Year B.Tech Project

Department of Computer Science and Engineering

Institute of Technical Education and Research (ITER)

Siksha 'O' Anusandhan (Deemed to be University)

Bhubaneswar, Odisha, India

Supervisor

Dr. Tusarkanta Dalai

Assistant Professor

Department of Computer Science and Engineering

ITER, SOA University

Disclaimer

This project was developed for academic and research purposes. Agricultural recommendations generated by the system should be verified by agricultural experts before field deployment.