uv-scripts/gliner
收藏资源简介:
GLiNER UV Scripts是一个用于在Hugging Face数据集上进行零样本命名实体识别(NER)的工具。它利用GLiNER模型,允许用户在运行时自定义实体类型列表,无需进行模型微调。该工具通过`extract-entities.py`脚本从文本列中提取实体,并生成包含实体信息的新的`entities`列,实体信息包括起始位置、结束位置、文本内容、标签和置信度分数。README提供了详细的使用指南,包括快速开始示例、在HF Jobs上运行的命令、从本地文件或存储桶读取数据的说明、推荐的实体类型词汇表、可用的模型选项、与Label Studio的配对使用以及一些注意事项和限制。
GLiNER UV Scripts is a tool for zero-shot named-entity recognition (NER) over Hugging Face datasets. It utilizes the GLiNER model, allowing users to pass a custom list of entity types at runtime without the need for fine-tuning. The tool extracts entities from a text column using the `extract-entities.py` script and generates a new `entities` column containing entity information, including start and end positions, text content, label, and confidence score. The README provides detailed usage instructions, including quick-start examples, commands for running on HF Jobs, instructions for reading data from local files or mounted buckets, recommended entity-type vocabularies, available model options, pairing with Label Studio, and some caveats and limitations.




