"The most perfect edition of plays ever published": the Digital Lacy project

Thomas Hailes Lacy (1809-1873)

Lacy was the leading theatrical publisher of "Acting Editions" -- practical working documents printed at 6d a copy for individual titles, or 5s for a bound volume of 15 titles.
Between 1848 and 1873, his Lacy's Acting Edition of Plays, grew to contain 100 volumes of 15 titles each: it was sold across the globe, and made him a reasonable fortune.
The LAE is a unique sample, apparently covering the full range of Victorian Theatrical presentations
The population it samples approximates to the titles listed in vols 4 and 5 of Allardyce Nicoll's magisterial History of English Drama -- c. 24,000 distinct titles performed between 1800 and 1900.

Research question: how representative is the LAE ?

A corpus is a sample, hopefully representative of a known population. Initial comparisons between the LAE and Allardyce Nicoll's Handlists suggest distributions of size, age, and mode are comparable.

"It is hard to avoid the conclusion that Lacy astutely leavened the mix for each volume, using mainly contemporary titles to complement the old favourites." (cf. How old are these plays?)

Digital Lacy project

Builds on and expands data from Richard Pearson's Victorian Plays Project (VPP), AHRB funded 2005-2007
- The VPP produced a catalogue of the LAE, along with c. 15,000 page images from a copy held at Birmingham Library
- These were OCRd, proofed, and made available in a PDF format as visually faithful as possible to the original
- By 2014, the project had processed 340 titles which were distributed from a site at the University of Galway
- Following Pearson's death in 2018, the project was frozen; by June 2022 the website had disappeared...
In 2022, with the aid of researchers who worked on the project, I recovered most of the resources it had created and transferred them to a github repository, where I continue to work on them
Digital Lacy now combines :
- a detailed and expanding set of metadata relating to the LAE and its authors, enhanced with links to available digital versions
- a slowly increasing number of TEI-XML transcripts

Proto-website at http://lb42.github.io/Lacy

Current workflow

Goal is consistent minimal encoding of a known source edition
VPP texts:
- VPP-PDF to Docx (OCR by Abby, thanks Huma-num)
- DocX to TEI-All (XSLT by TEI)
- TEI-All to Lacy XML (homegrown XSLT scripts)
Minimal markup, largely ignoring visual salience
TEI schema defined by ODD very close to dracor-schema

Impossible without manual intervention: this is the main bottleneck in current workflow.

DraCor vs Lacy: how close ?

DraCor and Lacy have a few ideological differences...

In DraCor metadata, the digital version is primary, any source version being nested within it; in Lacy, that hierarchy is reversed.
Some DraCor metadata (notably performances and identifiers) is relegated to a <standOff>; in Lacy it is imbricated in the TEI Header
DraCor uses explicit scene divisions to define stage-presence, as the basis for its network analysis; a quarter of Lacy titles don't have scene divisions.
Lacy uses many of the available TEI tags for the front matter of a play ; DraCor largely ignores these.
DraCor makes no attempt to support metadata such as editorial correction, variant readings, modifications for performance, etc; Lacy should, but doesn't.

However - the DraCor team is very responsive and helpful !

Tagging headaches are another persistent challenge

These texts are full of phenomena which break or strain the simple OHCO model...

speaker may be implicit or multiple
musical numbers (<spGrp>) don't tesselate and may self-nest
metadata may be missing, uncertain, inconsistent or just wrong

For example ....

Speeches assigned to multiple speakers

(<stage> not currently permitted within <speaker>)

Nesting of simultaneous speech or song

The whole dance (the Tyrolienne) is contained by a <spGrp> element which contains two nested <spGrp> elements, each containing two <sp> elements to be performed in parallel. (See also TEI Issue 2695)

Tentative suggestions and conclusions

TEI conformance is crucial to the interoperability of DraCor corpora. The DraCor profile/ODD should specify which parts of the TEI model are mandatory, desirable, permissible, unsupported ...
The documentation provided by the DraCor ODD is good, but could be improved: more examples and more discussion of edge cases would be useful ; as would simple tutorial guides showing how to use DraCor-conformant corpora with a variety of tools (not only python, plz)
Provide a forum for corpus creators to compare methods and tools, and to discuss possible solutions to common encoding problems
Encourage corpus creators to facilitate gap-filling in DraCor coverage, e.g. ECCO