bsnikhila/benchmark-test

NLU基准测试数据集是一个用于评估自然语言理解（NLU）模型性能的基准测试，主要包含两个任务：意图检测（Intent Detection）和命名实体识别（Named Entity Recognition, NER）。在意图检测任务中，数据集提供了带有黄金标签的查询，用于测试模型对用户意图的分类能力，主要评估指标为准确率（Accuracy），并同时报告宏平均F1分数（Macro F1）。在NER任务中，数据集包含分词后的句子及其对应的NER标签，用于评估模型识别文本中实体（如人名、地名等）的能力，主要评估指标为F1分数，并报告精确率（Precision）和召回率（Recall）。数据集结构包括两个配置：intent和ner，每个配置仅包含测试分割（test split），数据量在1K到10K之间（1K<n<10K）。数据集使用MIT许可证发布，语言为英语。The NLU Benchmark is a benchmark dataset for evaluating the performance of natural language understanding (NLU) models, focusing on two tasks: Intent Detection and Named Entity Recognition (NER). For Intent Detection, the dataset provides queries with gold labels to test the models ability to classify user intents, with the primary metric being Accuracy and also reporting Macro F1. For NER, the dataset includes tokenized sentences with NER tags to evaluate the models capability in recognizing entities (such as person names, locations, etc.) in text, with the primary metric being F1 and also reporting Precision and Recall. The dataset structure consists of two configurations: intent and ner, each containing only a test split, with data size between 1K and 10K (1K<n<10K). The dataset is released under the MIT license and is in English.