Saturday, 6 March 2021

Compositor: Fleuron 2.0

In March of 2017, I mentioned Fleuron: A Database of Eighteenth-Century Printers' Ornaments as a “new and enticing method for identifying” items printed by Thomas Gardner. (I also mentioned my three articles on the Gardner printing business, and have since used the post to aggregate newly-identified Gardner items.)

The woman behind Fleuron, Hazel Wilkinson, has continued "developing methods of identifying unknown printers using digital imaging" (i.e., using image-recognition software to identify and match printer’s ornaments). The upshot is Compositor, a site that went live in the middle of 2020—according to its blog and Twitter accounts—unbeknown to me. (Thankfully, David Levy got wind of the new site and recently let me in on the secret.††)

What Compositor enables is a lot closer to what I had hoped might be possible with an image-searching approach, than that provided by Fleuron. Fleuron was very good; Compositor is better. Although there is a lot of unnecessary, mouse-operated, busy work which makes any extensive search a grind, it is now possible to cross-match images of cast-metal Fleurons and carved wooden printers ornaments—Headpieces, Tailpieces, Initials, and Factotums—quickly compare the Compositor-identified image matches (to weed out the more obvious false-positives), and to trace the matches back to the works in which they appear.

So, for instance, I can [1] search Compositor for an item that I know was printed by Gardner, get [2] a full list of his ornaments that appear in that item, and then [3] search Compositor for other instances of each individual ornament, leading me to [4] items that I did not previously know were printed by Gardner—something I am keen to do.**

Likewise, I can [1] search Compositor for an edition of a work written by Eliza Haywood, follow steps [2] and [3], to [4] lead me to other items printed by the same printer as the printer of the Haywood edition. And since I do not know who the printer was of many editions of the works by Haywood, this should help me identify them—something I am keen to do.

Apart from the grind of unnecessary mouse-operated, busy work, which discourages anyone from undertaking the sort of extensive searches I am keen to have the results of, there are some other limitations worth mentioning.

Compositor appears to contain ornament images only from a sub-set of items filmed for The Eighteenth Century microfilm series, and subsequently scanned for ECCO (Eighteenth Century Collections Online). Also, the easiest way of locating an item on Compositor, is via an ESTC number. So, with a list of eighteenth century books in hand, you will be stymied if your candidate for a search

[1] the book isn’t on ESTC
[2] it is on ESTC, but not filmed for ECCO
[3] it is on ESTC, and on ECCO, but not on Compositor (!? I am not sure why this is the case)
[4] it is on ESTC, and on ECCO, and on Compositor, but the images captured by ECCOare so awful, that no match was possible
(*[5] it is on ESTC, and on ECCO, and on Compositor, but it has no ornaments! — obviously this is not the fault of Compositor, but it is easy to forget the basic, limiting, starting point for searches.)

It is a shame that Compositor does not allow you to upload your own images: even if you had to provide black-and-white images, of a certain number pixel density or size, include physical dimensions and so on, it would still be a huge improvement.

Even if Compositor only allowed you to source images from, say, the Internet Archive or Google Books, that would be an improvement, but I want to upload ornaments I have from other sources too, so I can identify the printer of such things as Sodom, or the Gentleman Instructed. A Comedy. By the E. of R. (Hague, Printed in the Year 1000000)—which is not on ESTC or ECCO or the Internet Archive or Google Books. The only known copy sold at Sotheby’s on 16 December 2004 and is now in private ownership.

Still, what Compositor does allow you to do is very impressive, and is a lot closer to what I imagined might be possible with an image-searching approach, when I explored getting such a project off the ground in 2004. If the search interface were to be improved (I might do a separate post on this) and if it were possible to search Compositor for user-sourced images, it would achieve all that I imagined in 2004.

Of course, it both changes were to be made, I would be unlikely to eat, drink or sleep until I had identified the printer of every item that survives from the first half of the eighteenth century. As things stand today, my main risk will be RSI. So, "Careful what you wish for"?

* * * * *

†† David has since done a post about Compositor here, giving an over-view of its use in his search for printers of works by or relating to Edmond Hoyle.

** Soon after writing the above I went ahead and did this search. I am not sure whether this blog is the best place to publish the full list of Gardner publications I have now identified—and my indecision is largely responsible for delaying the publication of the present post (which I wrote in January, about the same time as David was writing his post!).
[UPDATED 9 March 2021]

No comments: