GLiNER

https://arxiv.org/abs/2311.08526

A small but mighty model for information extraction ✨

Extracting information from documents such as invoices, grants and medical notes is probably the most popular use case of NLP. Small specialised models have an edge on this task since they offer better performance at a smaller cost 🚀 Large language models have mainly brought to the table the ability to extract information when no specialised model was available, i.e. out-of-the-box.

This out-of-the-box performance now comes at a smaller size with GLiNER 😮 GLiNER is a model trained on a large information extraction dataset annotated by a frontier LLM and filtered to contain only high quality examples. It’s zero shot (out-of-the-box) performance surpasses frontier LLMs which means you can develop and deploy an information extraction pipeline at a very small cost 💸 You can further improve that pipeline by collecting additional data and further training the model to your task.

🔗 Read more https://arxiv.org/pdf/2311.08526

/vide