Named Entity Recognition Explorer Tool

The NER Explorer Tool is an interactive web-based application designed to enable users to explore Named Entity Recognition (NER) in practice without requiring coding experience.The tool provides a user-friendly interface to access to four different NER models: spaCy's transformer-based model (en_core_web_trf)Flair's large English NER modelFlair's OntoNotes-based modelGLiNER's modern zero-shot model for custom entity types.Users can input text (optimal length: up to 5,000 characters), select a model, choose from 12 standard entity types (PERSON, ORGANIZATION, LOCATION, DATE, EVENT, FACILITY, GEOPOLITICAL ENTITY, LANGUAGE, NATIONALITIES/GROUPS, PRODUCT, WORK OF ART, MISCELLANEOUS), add custom entity types using zero-shot learning, and adjust confidence thresholds.The tool provides three types of output: Color-coded highlighted text showing identified entitiesSplit-color highlighting for entities identified by both standard NER models and custom GLiNER searches (marked with 🤝)Detailed tables with confidence scores and source attributionThis educational resource is specifically aimed at Digital Humanities (DH) researchers, Galleries, Libraries, Archives and Museums (GLAM) professionals, volunteers, and community professionals who wish to understand and experiment with NER methods for applications such as enhancing search and discovery in digital archives, automatic entity tagging, and allowing archival records to better 'speak for themselves'.The tool was created to: Address the difficulty of selecting appropriate NER modelsProvide access to models that performed best in testing on sample data during our projectTo raise awareness of zero-shot NER capabilities (GLiNER) compared to models with pre-defined entity types (spaCy, Flair),To demonstrate how traditional and zero-shot approaches can be used together.Built with Gradio, spaCy, Flair, GLiNER, and PyTorch, the tool runs entirely in a web browser and is deployed as a Hugging Face Space. It was created as part of the Digital Scholarship at Oxford (DiSc) funded research project "Extracting Keywords from Crowdsourced Collections."Acknowledgement This code to build this tool was developed with the assistance of Claude Opus 4 (Anthropic) Access https://huggingface.co/spaces/SorrelC/NER-Explorer-Tool