OpenVoiceOS/ovos-intent-template-bench-intents-for-eval
收藏资源简介:
该数据集是OVOS意图模板基准的预测数据集,专门用于评估意图分类模型。它基于OpenVoiceOS/intents-for-eval数据集,包含了OVOS Plugin Arena中模板范式意图联盟竞争者的每个样本预测结果。数据集按语言分割(如加泰罗尼亚语、丹麦语、德语、英语等),每个分割下提供JSONL文件,存储了竞争者的预测数据。这些数据遵循竞技场的标准合同,包括固定的数据集版本、插件版本、管道阶段以及精确匹配指标。数据集通过可复现的基准脚本生成,用于创建基准板、盲战池和基于基准的ELO阶梯。该项目由NGI0 Commons Fund和NLnet资助。
This dataset contains per-sample predictions for the template-paradigm intent league fighters of the OVOS Plugin Arena, based on the OpenVoiceOS/intents-for-eval dataset. It is designed for benchmarking intent classification models. The data is split by language (e.g., Catalan, Danish, German, English, etc.), with JSONL files for each competitor under predictions/<lang>/. Rows adhere to the arenas contract, including pinned dataset revision, plugin version, pipeline stage, and exact match metrics. Generated via a reproducible benchmark script, it supports the creation of benchmark boards, blind battle pools, and an ELO ladder. Funded by the NGI0 Commons Fund and NLnet.




