Interactive Grounded Language Understanding Dataset

本数据集名为‘Interactive Grounded Language Understanding Dataset’，由微软研究院等机构创建，旨在研究机器如何通过自然语言指令与人类协作完成任务。数据集包含8136条单轮交互数据，通过简化多轮对话为单轮交互，并利用在线众包平台收集。数据集内容包括自然语言指令、基于指令的行动以及澄清问题，主要应用于训练模型理解和生成自然语言指令，以及在指令不明确时请求澄清。

This dataset is named 'Interactive Grounded Language Understanding Dataset', created by Microsoft Research and other institutions. It aims to investigate how machines collaborate with humans to accomplish tasks through natural language instructions. The dataset includes 8,136 single-turn interaction samples, which are collected via online crowdsourcing platforms by simplifying multi-turn dialogues into single-turn interactions. The dataset covers natural language instructions, instruction-based actions and clarification questions. Its primary applications are training models to understand and generate natural language instructions, as well as soliciting clarifications when the provided instructions are ambiguous.

搜集汇总

背景与挑战

背景概述

本数据集由微软研究院等机构创建，旨在研究机器通过自然语言指令与人类协作完成任务，包含8136条单轮交互数据，涵盖自然语言指令、行动和澄清问题，主要用于训练模型理解和生成指令，并在指令不明确时请求澄清。