A general framework for natural language tasks for day-to-day usages using recent developments of ML and NLP. Specifically, transformers are crucially used here so the performance is quite good.
- Functionality
- Extension & server architecture
- Instructions for GPU/CUDA on server
- Basic NER functionality based on existing fine-tuned weights
- Better NER by fine-tuning on production-scale datasets
- Better entity recognition with wordpiece options
- Ability to use a remote server for NER
- User Interface
- Basic extension framework using React
- Basic content script to process NER on pages
- Summary view (all named entities mentioned)
- Better named entity display and user customizability
- Ability to jump to definition or link to unigraph
- Named Entity Recognition
- Identify named entities on webpages and replace them with links
