This is a report on an experiment to see if XML and the disaggregation of ingredient metadata could be used to reduce errors in recipes. Errors in web pages, PDFs, and print have been an irritant to authors, cooks, editors, and publishers for many decades, and occasionally the cause of an expensive recall. This research aims to see if markup could help.
Modern recipe structure is a well-established convention of a list of ingredients and a list of instructions (method). The writing about mistakes is sparse but highlights errors of omission and commission, inconsistency, sequence, and mismatch between the lists. Attributes for ID and classification were added to the ingredients list in a nonce recipe schema, along with an IDREFS attribute for use in the references to ingredients in the instructions; and code was written for cross-checking existence, usage, and consistent reproduction of names, and tested on a small collection of recipes.
We demonstrated that five of the seven classes of error identified could be straightforwardly remedied, but that the requirements for disaggregated data input needed to deal with the consistency issues may be too detailed for non-expert use unless assisted by a semantic filter.