PIZZA

PIZZA数据集是由亚马逊人工智能创建，用于解析比萨和饮料订单的任务导向解析。该数据集无法通过简单的插槽和意图表示其语义。数据集包含约245万条合成生成的语句，用于训练，而开发和测试集则较小，由人工生成。数据集提供三种版本：EXR表示法、TOP表示法和TOP-Decoupled表示法。PIZZA数据集旨在通过提供EXR表示来评估系统的端到端性能，并解决实体解析问题，同时克服TOP表示法的表达限制。该数据集适用于评估自然语言理解模型在处理复杂任务时的性能，特别是在需要结构化语义的场景中。

The PIZZA dataset was created by Amazon AI for task-oriented parsing of pizza and beverage orders. Its semantics cannot be captured by simple slot and intent representations. The dataset includes approximately 2.45 million synthetically generated utterances for training, while the development and test sets are smaller and manually generated. The dataset offers three variants: EXR notation, TOP notation, and TOP-Decoupled notation. The PIZZA dataset aims to evaluate the end-to-end performance of systems via EXR representations, address entity resolution issues, and overcome the expressive limitations of the TOP notation. This dataset is suitable for evaluating the performance of natural language understanding (NLU) models when handling complex tasks, particularly in scenarios requiring structured semantics.