A Recipe for Creating Multimodal Aligned Datasets for Sequential Tasks

Many high-level procedural tasks can be decomposed into sequences of instructions that vary in their order and choice of tools. In the cooking domain, the web offers many partially-overlapping text and video recipes (i.e. procedures) that describe how to make the same dish (i.e. high-level task). Aligning instructions for the same dish across different sources can yield descriptive visual explanations that are far richer semantically than conventional textual instructions, providing commonsense insight into how real-world procedures are structured. Learning to align these different instruction sets is challenging because: a) different recipes vary in their order of instructions and use of ingredients; and b) video instructions can be noisy and tend to contain far more information than text instructions. To address these challenges, we first use an unsupervised alignment algorithm that learns pairwise alignments between instructions of different recipes for the same dish. We then use a graph algorithm to derive a joint alignment between multiple text and multiple video recipes for the same dish. We release the MICROSOFT RESEARCH MUL-TIMODAL ALIGNED RECIPE CORPUS 1 containing ∼150K pairwise alignments between recipes across 4,262 dishes with rich commonsense information. * Work done when the author was an intern at Microsoft. 1 https://github.com/microsoft/ multimodal-aligned-recipe-corpus 7. Add 12 ounces of thawed peas and bean sprouts. 3. Add onion, garlic, peas and carrots. 4. Transfer shrimp to the hot skillet and cook them one minute per side.