Training Code and Dataset for AgriBioNER (ncRNA and Disease NER)

This repository contains the dataset, training code, and associated resources developed for ncRNA and Disease Named Entity Recognition (NER) in agricultural scientific literature as part of the AgriBioNER project. These materials support reproducible research and enable further development in agricultural biomedical text mining. 1. Annotated Dataset for Agricultural NER The dataset includes manually annotated agricultural research abstracts containing two specialized entity types: NON-CODING_RNA and DISEASE. All annotations follow a structured span-based format compatible with the spaCy NER framework. This dataset provides the foundation for training and comparing transformer-based and classical NER models within the agricultural domain. 2. Transformer Model Training Code (PubMedBERT + Other Transformers) The repository provides complete training scripts for PubMedBERT and all other transformer-based models evaluated in our study. Each model was trained using spaCy’s en_core_web_trf pipeline, ensuring a consistent architecture and reproducible training environment. The scripts cover data preprocessing, model configuration, training, validation, and export for deployment. 3. spaCy Baseline Model Training Code (en_core_web_lg) spaCy(en_core_web_lg) uses simple spacy pipeline without en_core_web_trf 4. AgriBioNER Web Application To demonstrate real-world usability, we provide a publicly accessible web-based application, AgriBioNER, which allows users to upload text or input agricultural abstracts for automatic ncRNA and disease NER. The application highlights extracted entities and provides interactive outputs such as tables, word clouds, and co-occurrence networks.Web App Link: https://huggingface.co/spaces/BhaveshKChoubisa/AgriBioNER