Cooking Recipes of the Middle Ages: Corpus, Analysis, Visualization

Cooking traditions, whether they are regional or in a larger context, are one of the most distinguishable items of European culture and an important part of European identities. But how did they become to what we know them now? How did they develop and what were their influences? During the last decades, research arrived at two important conclusions on these questions. Firstly, there are no quantitative studies on the origin and formation of regional cuisines in Europe. Secondly, manuscripts containing thousands of cookery recipes first appeared in the Middle Ages, which can be consequently regarded as the birth of modern European cuisines. On the European continent Latin, Middle French and Early New High German recipes provide the majority of culinary transmission. The project is preparing the cooking recipe transmission of France and the German speaking countries, which sums up to more than 80 manuscripts and about 8000 recipes, for the analysis of their origin, their relation, and their migration through Europe. The comparison of French and German food history is especially suited for this task as France always had a culturally formative influence on German speaking peoples.
Cooking recipes are culturally charged transient texts, which are best diachronically and spatially analyzed by strongly relying on digital humanities methods. However, understanding these recipes, their context and their transmission is not an easy process. The technical terms that describe ingredients, utensils, procedures, and customs of the time are a challenge even for history scholars who specialize on the topic. Thus, the texts need not only be transcribed and edited but also semantically enriched so that further analysis like machine aided comparison of ingredients or cooking processes can add to standardized philological research like the collation. The base of our data will be customized TEI/XML documents with a schema aiming at facilitating the semantic annotation of cooking recipes in general.
The core of our digital research strategy is the Semantic Web and the idea of Linked Open Data. We are in the comfortable position within the Humanities that we are mainly dealing with food ingredients i.e. animals, plants and fungi, all fields of research that are provided with a sophisticated amount of already established ontologies and Linked Data

For an overview of ontologies covering these topics see
http://www.ontobee.org/,
http://aims.fao.org/,
https://ndb.nal.usda.gov/ndb/,
https://agclass.nal.usda.gov/about.shtml,
http://zbw.eu/stw/version/latest/thsys/70498/about.de.html all of which are connected to the Linked Open Date Cloud by proving its data in one or more serializations of OWL and/or RDF(S).

including the general knowledge bases of Wikidata and DBPedia. Ontologies are also already being successfully used to represent (Dooley et al., 2018, Sam et al., 2014, Ribeiro et al., 2006) and analyze (Hammad and Hassouna, 2011, Vadivu and Waheeta Hopper, 2010) cooking recipes, albeit with different focuses and granularity of data. In the digital research-approach, we partly rely on text similarities but largely on the occurrence of ingredients, preparation instructions and time, tools, serving suggestions and medicinal, cultural as well as religious implications in the texts. The sheer quantity of semantics within one historic recipe shows how complex a semantic annotation of these texts is, leaving apart all philological observations that can be found in the material. Besides the ability to find links between resources that were not known before, a main argument for the decision to put Semantic Web technologies at the core of the project, was to ability to work outside of language barriers. Through the use of concepts in the sense of a notion, an idea rather than a term we are trying to overcome historical and language constraints i.e. the German word for potato ‘Erdapfel’ (‘erdaphel’ in Early New high German, literally ‘apple from the ground’) appears in a manuscript from about 1488) long before the potato was imported from South America to Europe, giving us proof that the concept of ‘Erdapfel’ must have been a very different one (probably ‘any kind of beet’) from today’s notion. For this reason, we needed to introduce a workflow that reflects the recipes’ philological and semantic complexity. A precompiled list of medieval plant names and their translations to modern English and German as well as their medieval variant dictions

http://medieval-plants.org
collected by Helmut W. Klug, the principal researcher of the Austrian part of CoReMa, gave us the opportunity to start with a (semi)automated alignment of Wikidata concepts with the Reconciliation Service API

https://github.com/OpenRefine/OpenRefine/wiki/Reconciliation-Service-API
provided by OpenRefine

http://openrefine.org/
. Once each term has a concept connected with it, these concepts are used to enrich the ingredients within the actual recipe texts in the TEI documents. However, as stated above, a crucial point of this process is the human interpretation of the enriched entities and the decision for a concept. Therefore, the auto alignment procedure can only be viewed as a means to an end.

Once the entities of each recipe are equipped with concepts, the project’s analysis can reveal concurring or deviating eating habits, text migration as well as the influence of neighboring countries on their respective cuisine. The vast implementation of ontologies in the natural sciences allows us to establish connections from historical eating habits to modern concepts of food and generate new knowledge for the domain of food history. The research data will also be the basis for spatial and temporal visualization and statistical evaluation. The storage, analysis and dissemination of the project’s data is handled by the data repository GAMS (Geisteswissenschaftliches Asset Management) developed by the Austrian Centre for Information Modelling in Graz.

http://gams.uni-graz.at/
Within this infrastructure aimed at long-term preservation, the triplestore Blazegraph is accessed through a web service for the storage and retrieval of RDF triples which allows us to query our project’s databases of medieval recipes in France and Austria as well as all connected concepts in the Linked Open Data Cloud.

Bibliography

Dooley, D. M., Griffiths, E. J., Gosal, G. S., Buttigieg, P. L., Hoehndorf, R., Lange, M. C., Schriml, L. M., Brinkman, F. S. L. and Hsiao, W. W. L. (2018). FoodOn: a harmonized food ontology to increase global food traceability, quality control and data integration.
Npj Science of Food,
2(1): 23 doi:
10.1038/s41538-018-0032-6.

Hammad, R. and Hassouna, M. (2011). Multi-Language Semantic Search Engine. : 6.

Ribeiro, R., Batista, F., Pardal, J. P., Mamede, N. J. and Pinto, H. S. (2006). Cooking an Ontology. In Euzenat, J. and Domingue, J. (eds),
Artificial Intelligence: Methodology, Systems, and Applications. (Lecture Notes in Computer Science). Springer Berlin Heidelberg, pp. 213–21.

Sam, M., Krisnadhi, A. A., Wang, C., Gallagher, J. and Hitzler, P. (2014).
An Ontology Design Pattern for Cooking Recipes – Classroom Created.

Vadivu, G. and Hopper, S. W. (2010). Semantic Linking and Querying of Natural Food, Chemicals and Diseases.
International Journal of Computer Applications,
11(4): 35–38 doi:
10.5120/1567-2093.