The London Brewers' Prosopography: Parsing Historical Apprenticeship Records Harvey Quamen hquamen@ualberta.ca University of Alberta, Canada This poster describes one preliminary aspect of a new project about the history of beer and brewing in London. I am building a prosopography of brewers' apprentices in the years from about 1530 to approximately 1800. A catalogue of apprentices and their masters can teach us not only about the social and cultural history of British beer during that time but about the social network of people working in the industry. During this 270-year period, the Worshipful Company of Brewers, the medieval brewing guild first established by Henry VI in 1438, logged nearly 10,000 apprenticeship records (Webb), a collection that serves as just one of many potential datasets that can yield insights into England's brewing culture. The more immediate goal described in this poster, then, is how to parse these 10,000 records into component parts— people, places, occupations, and dates—so that these relationships can be analyzed and mapped over time. A typical apprenticeship record looks something like this: AMBROSE John s John, JlsLey, I3rk, xn^JX§i6Tto Samuel May 14 Jul 1703 In other words: John Ambrose, whose father is John Ambrose from Ilsley, Berkshire and is a maltster by profession, was apprenticed to Samuel May and the fee was paid on 14 July 1703. The record lists three people, a parish, a county, a profession, and a date—a typical dataset found when tracking apprenticeships (Lane). The simplicity and regularity of that template tantalizingly suggests writing an automated parser, which I am doing with an open source Python module called pyparsing. Although these recursive descent parsers, as they are called, are designed for more elaborate projects (like writing compilers), they are the perfect tool for a job like this because they allow users to construct grammatical “rules” that look like simple additions. For example, the record above can be parsed according a grammatical rule that looks like simple Python: apprentice + father + location + occupation + to + master + date But the wide variety of apprenticeship records presents a challenge. As it turns out, the first 10% of the apprenticeship records require nearly 40 diff erent template “grammars.” The effectiveness of an automated parsing approach—useful but nonetheless somewhat limited—is the main point of the poster. Supplementary strategies (like natural entity parsing and dictionary lookups) may provide some help and, if they prove themselves worthy, they will become part of the presentation as well.