Last week, EMROC published a blog post called “Code Breakers,” describing the efforts of our Before “Farm to Table” project volunteers, who – like EMROC – transcribe our recipe books. This week’s blog post will be talking about our “Code Makers”: Meaghan Brown (Digital Production Editor), Michael Poston (Database Associate), and myself, Elisa Tersigni (Digital Research Fellow), who together devised BFT’s process of encoding—or what happens after the transcriptions are vetted and before they become accessible online. This blog post is meant to be a crash-course in what this in-between stage looks like and why we do it.
What is encoding?
Encoding is the process of applying code to digital texts to create machine-readable texts in order to facilitate computer-assisted analysis. Our encoding is done using Extensible Markup Language (XML), which is commonly used because it is intuitive and flexible. (More on this below.) Encoding might be thought of as functioning the same way that highlighting in a text does. Take, for example, this recipe for “An excellent medicine For any kinde of Ague”:
Let’s say we wanted to encode the recipe’s text to identify easily the title (which we will mark in red), all ingredients (which we will mark in yellow), any person’s names (in blue), and any markers of approval or disapproval of the recipe (green). The colour-marked recipe will look like this:
An XML-encoded version of this recipe might not look much different, except phrases are sandwiched between a pair of tags (a start tag, which marks the beginning of the phrase, and an end tag, which marks the end of the phrase) instead of being highlighted:
<head>An excellent medecine For
any kinde of Ague.</head> <persName>Doctor Costine</persName>
Take <ingredient>saffron</ingredient> one ounce and a halfe and
as many <ingredient>Currens</ingredient> vnwashed as will
beate itt up into a Cataplasme, beat
them well together and putt itt
in bagges of Poulter two to be
applyed one the hand wristes one the
Pulses and the other one the pitt
of the stomach: <tested approved= “yes”>probatum est</tested>
(N.B.: The above tags are for illustrative purposes only and may not reflect actual encoding.)
We use XML because it is both human readable and machine readable: the human reader can intuit that a phrase falling in between the “ingredient” tags is an ingredient. Our encoding is based on the Text Encoding Initiative (TEI)’s Guidelines, which is an encoding standard that is commonly used for digital humanities projects. TEI is a consortium that collectively develops and maintains encoding standards for texts. Standards are important to ensuring that data is easily shareable and usable by other researchers for other projects. Standards also help with preservation of data, making sure that the data is useable for as long as possible.
Encoding permits computer-assisted reading. For instance, if all the recipe titles in a book are marked as “head”, we can generate a list of all recipe titles by extracting everything between the “head” tags. If recipe titles are marked in 20 different books, we can generate a single list of all recipe titles, or we can generate 20 lists of recipe titles and compare those lists against one another to see if there is any overlap between books. Or, if we mark the recipes as having some evidence of being tested and include details of whether the recipe book user expressed approval or disapproval of the recipe, we can extract all the recipes of which users approved or disapproved.
Encoding doesn’t just help users perform searches; it can also help software development. For example, if—instead of marking titles, ingredients, and people’s names—we instead marked the structure of the manuscript—line beginnings and endings, underlining, formatting—high-quality encoded transcriptions can be used to create or improve Handwritten Text Recognition (HTR) software for manuscripts.
What are we encoding?
Because encoding requires interpretation, it is time consuming. And because encoding is flexible, it is prone to what is called ‘scope creep’—that is, the tendency for expectations and demands to increase over time, thereby slowing down a project. Our encoding must balance breadth (the number of manuscripts we encode) with depth (the amount of encoding we complete in one text). To strike that balance, our project began with a trial period, in which we tested how long it would take us to encode certain features, such as including both original and modern spellings in transcriptions.
Our trials helped us decide to prioritize the identification of the basic structure and details of the texts: where recipes start and end, the names of people identified in recipes, the names of locations identified in recipes, and whether recipes were tested and approved. These details will help us to do some basic network analysis to reveal some of the relationships between our recipe books: who owned them, where they came from, and whether recipes are shared across time and space.
For example, by searching the term “lady” in the Folger’s Luna Digital Image Collection, which currently has searchable text versions of some of our manuscript books, I was able to discover that variations of “Lady Allen’s water” appear in at least half a dozen of our manuscript recipe books from the 17th century: Jane Staveley’s receipt book (V.a.401), Susanna Packe’s receipt book (V.a.215), Lady Grace Castleton’s receipt book (V.a.600), and two anonymous books: V.a.563 and V.b.363. This phrase does not currently appear in the Oxford English Dictionary (OED) and appears only twice in print, according to searches of the Early English Books Online (EEBO) database. The ability to search quickly not only helps answer research questions but can also point researchers in to new research questions: what is “Lady Allen’s water,” from where did the recipe originate, and why is it commonly recognized in manuscript receipt books but not in print recipe books?
How can I learn more?
If you’re a graduate student who is interested in our project, consider applying for our funded graduate student workshop, Eating Through the Archives: Interdisciplinary Approaches to Early Modern Foodways. If you’re interested in volunteering for our program as a Code Breaking transcriber or a Code Making encoder, contact Heather Wolfe at hwolfe [at] folger.edu. You can also read our research updates by following @FolgerLibrary or @FolgerResearch on Twitter or reading our scholarly blogs The Collation and Shakespeare & Beyond.
Elisa Tersigni is the Digital Research Fellow for Before ‘Farm to Table’: Early Modern Foodways and Cultures, Folger Shakespeare Library. Elisa tweets as @elisatersigni.