Microsoft Research Multimodal Aligned Recipe Corpus

为了构建MICROSOFT RESEARCH多模态对齐配方语料库，作者首先从网络中提取了大量文本和视频配方。目标是找到同一道菜的多个文本食谱和多个视频食谱之间的联合比对。这项任务是具有挑战性的，因为不同的食谱在他们的指示和使用成分的顺序上有所不同。此外，视频指令可能是嘈杂的，并且文本和视频指令在其描述中包括不同级别的特异性。

To construct the Microsoft Research Multimodal Alignment Recipe Corpus, the authors first extracted a large number of text and video recipes from the web. The core objective is to identify joint alignments between multiple text recipes and multiple video recipes for the same dish. This task is challenging, as different recipes vary in their instructional content and the sequential order of ingredient usage. Furthermore, video instructions may be noisy, and both text and video instructions include different levels of specificity in their descriptions.

搜集汇总

背景与挑战

背景概述

该数据集由Salesforce Research和Microsoft Research于2020年发布，旨在构建一个多模态对齐的食谱语料库，通过从网络提取文本和视频食谱，实现同一道菜的不同食谱之间的联合对齐。由于食谱在指令、成分顺序和描述特异性上存在差异，这项任务具有挑战性。