Abstract
Food recipes, as a type of instructional text, contain series of instructions that guide human users\r through cooking processes. Their structures can usually be represented as flow graphs: at the start\r state, a collection of ingredients are introduced. Then, a series of actions are performed on these\r ingredients (along with cooking tools). Finally, the completed dish is produced at the end.\r A computer system that can extract such structures from food recipes can have many useful\r applications.\r Since food recipes are written by humans, they pose many of the same problems the field of\r Natural Language Processing is trying to solve. Furthermore, they also have challenges that are\r not commonly seen in many other types of texts.\r This thesis explores a Statistical Machine Learning approach to food recipe text processing. I\r will detail my work in the creation of an annotated food recipe dataset used in supervised learning.\r I will also describe my work in training Machine Learning algorithms that perform Named Entity\r Recognition and Relation Extraction tasks on this dataset.