YnJhY2lzMjAyNnRleHQyc3Fs/environmental_registry_test_set
收藏资源简介:
该数据集是一个匿名的基准测试集,用于评估基于真实PostgreSQL/PostGIS环境注册数据库的葡萄牙语文本到SQL的代理性能。数据集包含180行数据,分为测试集,其中包含90个SQL问题、45个澄清问题和45个不可回答问题。SQL问题的难度分为20个简单、30个中等、30个困难和10个专家级别。数据集使用巴西葡萄牙语,数据库类型为PostgreSQL/PostGIS环境注册数据库,主要模式包含19个表,涵盖农村财产、市政区、土著土地、保护单位、环境禁运、定居点、quilombola领土、联邦公共土地、森林砍伐事件和空间重叠事实等。数据集还包含了评估设计、报告结果和许可证信息。
This dataset is the anonymized primary benchmark used for evaluating agentic Portuguese Text-to-SQL over a real PostgreSQL/PostGIS environmental-registry database. The dataset contains 180 rows, split into a test set with 90 SQL questions, 45 clarification questions, and 45 unanswerable questions. The SQL questions are divided into 20 easy, 30 medium, 30 hard, and 10 expert levels. The dataset is in Brazilian Portuguese, and the database type is a PostgreSQL/PostGIS environmental-registry database, with a primary schema containing 19 tables covering rural properties, municipalities, Indigenous Lands, conservation units, environmental embargoes, settlements, quilombola territories, federal public lands, deforestation events, and spatial-overlap facts. The dataset also includes evaluation design, reported results, and license information.




