MDID

该数据集名为MDID，包含了1299条Instagram帖子，这些帖子被标注了作者的意图以及图像与文本之间的情境和符号关系。它是一个多类别数据集，每个样本都根据现有的修辞类别被标注了8个意图标签。该数据集专注于探究作者意图以及图像与文本之间的关系。规模上，它包含了1299条帖子，任务是对意图进行多类别分类识别。

Named MDID, this dataset includes 1299 Instagram posts annotated with the authors' intentions as well as the contextual and symbolic relationships between images and their accompanying texts. It is a multi-class dataset, where each sample is assigned 8 intent labels according to existing rhetorical categories. This dataset focuses on exploring authors' intentions and the intermodal relationship between images and texts. With a total of 1299 posts, its core task is multi-class classification for intent recognition.

搜集汇总

背景与挑战

背景概述

该数据集是用于多模态文档意图分析的Instagram帖子数据集，支持文本和图像特征的整合研究。数据集包含训练和验证分割，提供ResNet-18图像特征和标注信息（如意图、符号学、上下文类别），但未包含原始图像。它旨在帮助研究者复现论文结果并进行多模态分类实验。