The Reappearances of St. Basil the Great
in British Library MS Cotton Otho B. x
Kevin Kiernan, Brent Seales, and James Griffioen
One of St. Basil the Great's great miracles was rendering a manuscript totally illegible. He was able to ruin all but one reading in this manuscript during his lifetime, and he obliterated the last stubborn reading as his first posthumous miracle. While he was still alive, he used "remote access" in another miracle to alienate a manuscript from its rightful owner (the devil, as it happens), then tore it up. It seems therefore devilishly ironic that the Old English versions of his life mostly survive in ruined bits and illegible pieces -- including two halves of the folio describing the manuscript he ripped apart! In an interdisciplinary project called "The Digital Atheneum: New techniques for restoring, searching, and editing humanities collections," we are attempting to reverse some of the damage done to manuscripts in the infamous Cottonian Library fire of 1731, and to make them accessible again through electronic editions.1 We chose to confine our research to heavily damaged manuscripts, because we would presumably master easy problems while striving to solve difficult ones, and we would make available some currently unusable material for new research. In this group of manuscripts one of the biggest challenges for Anglo-Saxon scholars and computer scientists alike is Cotton Otho B. x, once a large Old English miscellany dominated by saints' lives written by an eleventh-century monk named Ælfric. The first saint's life in this wreck of a manuscript is the Life of St. Basil the Great, which furnishes many opportunities to show how new computing techniques and electronic editions can help restore damaged manuscripts and provide easy access to formerly inaccessible texts.2
The Anglo-Saxons of the tenth and eleventh centuries venerated the prolific Church Father, St. Basil the Great (330-379).3 To judge by the Life he wrote, Ælfric admired St. Basil as an activist bishop and monk who wrote a monastic rule, contributed to the Eastern ritual of the Mass, fought heresies, bravely stood up to emperors, and worked many miracles in the service of his flock. In the early tenth century King Athelstan (924-939), a spectacularly successful collector of saints' relics, had acquired among his hundreds of relics one of St. Basil's teeth as well as his bishop's crosier.4 In addition to the extant reliquary lists, evidence of his cult in late Anglo-Saxon England comes from seven surviving liturgical calendars5 as well as from three extant vernacular copies of his life.6 Two of these manuscripts of his life are in fragmentary state, eerily reminiscent of that reliquary tooth King Athelstan acquired, but one full version survives in the handsomely written and well-preserved British Library MS Cotton Julius E. vii, the basis of our only modern edition.7 The complete Julius version allows us to understand where the other surviving fragments once belonged in their respective manuscripts. The University of Toronto's Dictionary of Old English project has greatly simplified the task by providing an online version of Skeat's edition.8 It is as if King Athelstan had St. Basil's dental records to go with his reliquary tooth.
Even with the aid of Skeat's edition of St. Basil, however, it is impossible for a reader today to sit down with Cotton Otho B. x and read the remaining fragments in their correct sequence. The manuscript was in such miserable condition after the fire of 1731 that it was considered unusable and was packed away and forgotten for well over a century in a garret in the British Museum. In 1836 Sir Frederic Madden rediscovered it and many other lost manuscripts in the garret, and as Keeper of Manuscripts assigned a staff member named N.E.S.A. Hamilton to try to sort out the thoroughly disordered leaves in preparation for restoration binding. Madden was not satisfied with the results and subsequently tried to order the leaves properly himself, but without great success.9 For example, after all of Hamilton's and Madden's efforts, one must still today be ready to flip pages back and forth to read the first four folios of Cotton Otho B. x in their correct order: first, fol. 60, beginning with the verso; second, fol. 36, also beginning with the verso; third, fol. 49; and finally fol. 1.
To read the St. Basil portion in sequence is no easy task. One must begin with folio 3, skip to folio 5v, and then move back to folio 4. After that, to stay in sequence, the reader must take a day-trip to Oxford and the Bodleian Library to study a stray leaf from Cotton Otho B. x that was taken there in November 1731, a couple of weeks after the fire.10 The reader must stay alert, because the Bodleian Library, in the tradition of this unlucky manuscript, has mislabeled the leaf with the verso as the recto, and the recto as the verso. Finally (perhaps more taxing than the trip to Oxford), this diligent student must return to London and the manuscript in the British Library, and try to read line by line from fragment 50 to fragment 6 throughout the recto, and line by line from fragment 6 to fragment 50 throughout the verso, because "folios" 50 and 6 in the manuscript are actually two parts of the same folio in the wrong order. In short, the texts in Cotton Otho B. x are practically inaccessible and are accordingly ideal candidates for some radical restoration.
The official foliation is not only useless, but actually detrimental to a study of the manuscript, because its numbers lead readers in all the wrong directions. To facilitate study of this manuscript Kiernan has devised an "electronic foliation" that presents the leaves in their correct order, but also keeps track (in parentheses) of their official British Library foliation numbers and of the Bodleian Library pressmark for MS Rawlinson Q. e. 20, the leaf from St. Basil now alienated in Oxford. With all the surviving leaves of Cotton Otho B. x in order, St. Basil should begin on the fifth folio of the codex, not the third. Thus for the five (not really six) surviving leaves of St. Basil this e-foliation is folio 5(3), 6(5v)r, 6(5r)v, 7(4), 8(RQe2Ov)r, 8(RQe2Or)v, 9(50+6) for the recto, and 9(6+50) for the verso. Here is an overview of all the surviving folios:
This useful e-foliation will never have more than a virtual reality, for two overpowering reasons: although they may easily rectify the mislabeling of the recto and verso, the Bodleian Library, Oxford, is not likely to repatriate the leaf it now owns to its original manuscript at the British Library; and the British Library, even if it does eventually put the leaves of Cotton Otho B. x into their proper order, is not likely to remove fragments 50 and 6, the misbound parts of a single folio, from their protective paper frames and rebind them together.11 With the cooperation of the respective repositories, however, these highly improbable events are of course easy to accomplish in an electronic edition.
By putting together all of the extant fragments of the Life of St. Basil in Otho B. x and fully collating them for the first time with the version that survives in Julius E. vii, it is possible to see how much survives of the original manuscript. The best preserved leaves show that there were 29 lines of text per folio page. The fragment of a leaf at Oxford preserves only about 22 lines, but we have the text from Julius to show what was lost at the side and the bottom of the recto, and we can restore with relative certainty the amount that was similarly lost at the side and the bottom of the verso. Between the end of fol. 8(RQe2Or)v and the beginning of the split leaf, fol. 9(50+6)r, are 66 lines of verse, which would have fit on a single folio (58 MS lines). Thus nearly half of the original manuscript of the Life of St. Basil the Great in Cotton Otho B. x still survives today.
One manuscript leaf that might have been a fascinating research project is unfortunately beyond restoration, because St. Basil destroyed it. In the culminating episode of Ælfric's Life of St. Basil, a rich widow who had been living "as a pig in muck" (swa swin on meoxe Skeat 528) determines to amend her life by writing down all her sins on a vellum leaf, sealing it with lead, and then asking St. Basil to obliterate the list of sins. Deciding to take on this innovative approach to penitence, St. Basil prays that as Christ's own deed blots out sins and as all our sins are written down with Him, Christ should help him deface the manuscript (Skeat 541-548). After a night of prayer, St. Basil successfully blots out all but the most mortal sin (Skeat 551-553). Claiming he is too sinful to eradicate this one, he sends her to a hermit in the wilderness to erase it. The long journey seems a ploy to get rid of her when St. Basil, having foreseen the time of his death, immediately prepares to die as soon as she leaves.12 It is a curious excursion, because her pilgrimage turns out to be a waste of time. The hermit tells her that only St. Basil the Great can help her!13 Rushing back to Caesarea, she is understandably upset to find St. Basil dead and lying on his bier. "She then threw the writing on the bier, and told the men about her misdeeds." One of the attending priests, curious about her remaining sin, retrieves the manuscript and sounds frustrated when he discovers that everything is obliterated. "Why are you so worked up, woman?" he yells at her. "This vellum is blotted out!" (Skeat 640-644).
What does this strange story have to do with restoring manuscripts? A modem-day miracle would satisfy our desire to know the past and make the writing legible again. Ultraviolet and digital image-processing can to some extent gratify these mundane desires on many of the illegible leaves that have survived from St. Basil.14 The first remaining folio of Cotton Otho B. x, fol. 5(3)r, is a good example. It resembles a medieval Hell's mouth that has devoured the text the missing vellum was supposed to save. Collation with the surviving version in Cotton Julius E. vii reveals that both the righteous and the unrighteous have been lost in this seemingly ravenous maw, including Eubolus, St. Basil's teacher and subsequent disciple, other learned Athenian philosophers, and two distinguished classmates, St. Gregory the Theologian of Nazianzus, and the Roman and Byzantine emperor Julian the Apostate (360-363). Historians (wyrdwriteres) and their books (bocum) have not fared well in this Hell's mouth, in addition to the philosophers, theologians, and emperors.
Moreover, charring, discoloration, new layers of tape, and relatively recent attempts at preservation by applying gauze over glue, which has turned opaque over the years, have made it very difficult to read much of the text that has survived around the gaping hole. Fortunately, ultraviolet counteracts some of these features obscuring the text. By digitizing the ultraviolet effects and then processing the images, we can clearly restore, for example, the badly faded rubric and even penetrate the gauze and glue.
Even in strong light the leaf is so severely charred that it is difficult to read anything, but ultraviolet and contrast enhancement show that the rubric reads K[a]l[end] IAN[uarii] DEPOSITIO S[ancti] BASILII, attesting to Ælfric's eccentric dating in his sanctorale of St. Basil's feast day on 1 January, the date of his burial, instead of 14 June. Beneath the gauze to the left in line 5, it is now possible to see the normal genitive plural reading wintra, instead of the variant spelling wintre that appears in the Cotton Julius E. vii manuscript. Ultraviolet and image processing also disclose the scribe's superscript correction of haten to gehaten in the first line, a reading that scholars after the fire have been unable to see through the scorched vellum.15 Here the scribe's superscript ge is digitally highlighted between wæs and haten:
Similar digital techniques reveal scores of variant readings Skeat was unable to detect and record in his published collations. Skeat says that he gives all variations that he is able to decipher,16 but it is clear from studying the ultraviolet images that he was unable to see the great majority of variations. For example, in one twelve-line stretch of text on fol. 7(4)v3-15 (lines 172-185) Skeat finds only one variant reading, segene for sægne, "a statement, saying," whereas the ultraviolet image reveals over two-dozen variants.
Another episode in St. Basil's career may be used to introduce the topic of remote access. The story is told on fol. 8(RQe2Ovr)rv, the leaf now in the Bodleian Library, Oxford (Skeat 204-264). After an arrogant altercation with St. Basil stemming from their schooldays in Athens, Julian the Apostate, who is on his way to fight the Persians, promises to lay waste to Caesarea when he returns. St. Basil warns the citizens of Julian's rage, and advises them to raise a tribute to placate him, advice that would have resonated with an Anglo-Saxon audience during the Viking incursions. Enjoying his own remote access, however, St. Basil is visited that night by the Virgin Mary and a heavenly host. They promptly mobilize the martyred St. Mercurius, whose relics and armor lie in St. Basil's church, to execute Julian in his camp for denying Christ and generally speaking pompously.17 The Bodleian Library, Oxford, has recently made this leaf from Cotton Otho B. x accessible by remote access through its splendid website of manuscript images.18 It is not apparent from the online images that the leaf is preserved in its own reliquary, sealed between glass plates, and kept in a small, book-like, solander made for the purpose. What can be seen, sealed with the fragment, is a note in the handwriting of the eighteenth-century Oxford antiquary, Thomas Hearne, explaining how it got into Oxford: "A Fragment of some MS. that suffered in the Loss by fire of the Cotton Library. Given me by Browne Willis, Esq. being brought to me by his son a commoner of Xt ch [i.e. Christ Church] Nov. 15. 1731."
In the context of digital restoration, it is interesting to compare the different ways that the British Museum and the Bodleian Library restored these fragments from the Life of St. Basil in the nineteenth century. At the British Museum Henry Gough had perfected the process of inlaying leaves in paper frames, a process well-known from the Beowulf manuscript, and one that was used with varying success with the two damaged manuscripts containing the Life of St. Basil. Ironically, Gough began working at the Bodleian, but was hired by Sir Frederic Madden to undertake the massive task of restoring the most ruined Cottonian manuscripts.19 Gough's method was to trace each leaf on heavy construction paper, cut out the center leaving a retaining edge, and then paste the vellum leaf into the open space. One disastrous problem in the case of Otho B. x is that he used the then new acidic paper for the frames, which have begun to crumble and have leached out stains from the leaves. A.S. Napier records the method of restoration for the Bodleian leaf in his 1887 note, "A Fragment of Ælfric's Lives of Saints," published a few months after the leaf was found by a librarian in a drawer in the Bodleian Library. Napier relates that "it was wrapped up in a piece of paper" containing the previously mentioned note by Thomas Hearne (378). According to Napier, "the fragment itself was so shriveled up and blackened by the heat that it was quite impossible to decipher it until it had been soaked in water and carefully stretched." The smooth edges of the fragment indicate that some trimming, whether on purpose or by accident, was done before the fragment was, as Napier says, "laid between two pieces of glass." The piece of paper that kept it safe from 1731 to 1887 was certainly trimmed, preserving only Hearne's note:
This front view of the leaf is, as we have seen, mislabeled as the recto, rather than the verso. From what can now be deciphered in strong daylight or in the digital image, it appears that the text, especially on the recto, is no longer as legible as it was when Napier transcribed it. Based on Kiernan's experience with the other Cotton Otho B. x leaves, the text would almost certainly respond well to ultraviolet, but whether ultraviolet would penetrate the glass is perhaps less likely. In any case, the glass casing is not a good way to preserve the natural suppleness of vellum, and the Bodleian Library might consider using modern conservation techniques for the leaf, as the British Library is doing for the rest of the manuscript.20
The next surviving leaf is now mounted as two separate folios in Cotton Otho B. x, the widely displaced fragment 50 and its other half, the correctly situated fragment 6, which I have digitally rejoined as one leaf and named 9(50+6)r and 9(6+50)v. The projecting vellum at the bottom of fragment 6 has shrunk inward, which prevents bringing the two fragments more closely together without obscuring some text on the adjacent fragment.
It is perhaps significant that the text missing from the Bodleian leaf, which we can restore from Cotton Julius E. vii, has the same L-shaped form as fragment 50. The text on fragment 6, moreover, breaks off in the same area as the Bodleian leaf, around line 22. It may be that these leaves were cut for some unknown reason at the same time, perhaps during efforts to extinguish the fire.
By strange coincidence, St. Basil himself tore apart a manuscript the devil acquired on this same leaf, fol. 9(6+50)v14-18 (Skeat 379-383). The Faustian story on this reunited folio is about a young man who makes a pact with the devil in order to marry the girl he sinfully loves. The devil dictates to the youth a contract renouncing Christ and his baptism, assuring the devil, he thinks, of the boy's company on Doomsday. After the boy and girl marry, in a part of the Life preserved only in Cotton Julius E. vii, St. Basil fights with the devil on behalf of the repentant young man to retrieve the autograph manuscript. St. Basil is unimpressed with the legality of the document, and as a result of his prayers the contract falls from the ether into his hands. After he confirms with the youth that it is indeed his handwriting, St. Basil promptly tears it up (Skeat 458). Skeat has memorably illustrated how difficult it is to read fol. 9(50+6)r and 9(6+50)v with the two parts so widely separated. He realized that the text came from about the same place in the story, but mistakenly concluded that fragment 50 must be from a different manuscript of the same text as fragment 6: "As noted at p. 70," he says, "one of the leaves in this MS. (leaf 50) does not belong to the MS. at all, so that the collations are here marked with the symbol O2." According to Skeat, "It is easy to see whence the leaf came, viz. from the other much burnt Cotton MS. with similar contents, i.e. from MS. V (Vitellius D. 17)" (p. xvi).21
In fact it is easy to see from the script and layout that Otho B. x and Vitellius D. xvii are quite different manuscripts, and N.R. Ker readily saw that fragments 50 and 6 were two parts of the same leaf from Otho B. x.22 Skeat's blunder nonetheless shows how difficult it is to read the two parts of the same folio when they are separated by 44 folios. Even if they are ever rebound in the correct order, these two parts will most likely remain in their paper frames, because removing them might cause further damage. A digital restoration in a single image, on the other hand, is easily accomplished, once or as many times as desired, without any possible harm to the manuscript.
The fragments are reasonably legible, even in ordinary light, and bringing them together and enhancing the contrast render them the most easily read of all the surviving leaves. A "free transform" procedure in Photoshop makes it possible to join the fragments more closely, by moving to the right the bit of shrunken vellum on fragment 6 that would otherwise cover some text on fragment 50. The same free transform function could bring the two fragments quite close together.
It is important to keep in mind, however, that these fragments are not really planar. The digital camera produces a flat, two-dimensional result, which is an accurate facsimile only if the original object is also flat. In fact, each manuscript leaf in Cotton Otho B. x has its own three-dimensional properties, which appear as ambiguous distortions in the two-dimensional digital photograph. These three-dimensional properties in the objects themselves were not intentionally created but were caused by the way they bend when the manuscript is opened, by their different texture caused in part by the action of fire, water, and shrinkage, and by the way each has reacted to the individual paper frames that hold them.23
University of Kentucky computer scientist Brent Seales and his doctoral student Michael Brown are currently experimenting with three-dimensional modeling to see if they can more accurately rejoin the fragments by taking into account the constantly changing three-dimensional properties of each leaf. In order to record the structure of a three-dimensional object, Seales and Brown have built an inexpensive, portable device from commonly available off-the-shelf hardware, a digital light, or LCD (liquid crystal display), projector. With this device, one can capture millions of three-dimensional sample points which together produce a very fine reconstruction of the shape of the surface of the fragments. These points form the basis for a mesh of triangles that approximates to a very fine degree the shape of the fragment, and onto which the high-resolution texture from the digital photograph is rendered to give an accurate, metric rendition of the shape and color of the object. Complete shape information together with high-resolution digital photography make it possible to view the mesh as a textured image looking very much like a leaf in the manuscript, and also to view it as a "wire-frame" mesh, the structure of triangles from the points recording the three-dimensional properties of the object.
Seales explains that, while there are many technologies available for three-dimensional acquisition, four important considerations led him to design the system we are using for three-dimensional scanning of the damaged manuscripts in the Cottonian collection:
Specialized hardware for acquiring three-dimensional images is very expensive;
Most existing systems are not designed for imaging the same object under a variety of lighting conditions;
The use of a laser is often discouraged for real or perceived hazards with sensitive materials;
Most existing systems are not easily integrated with available digital camera setups.
An important design and deployment goal was to acquire three-dimensional shape representations without expensive alterations of the digital camera setup in the British Library. Also, as we knew that fiber-optic backlighting and ultraviolet fluorescence often restored readings that were invisible in ordinary lighting, it was important to use a system that allowed the recovered three-dimensional data to easily register with imagery under different lighting conditions. Furthermore, not all artifacts require three-dimensional scanning, and in fact two-dimensional imaging normally suffices for the relatively flat surfaces of the damaged manuscripts. Considering these factors, Seales decided that laser scanners or similar devices were inappropriate for our purposes.
There are two primary reasons why highly collimated light sources, or lasers, are problematic. The first is technical and the second psychological. The technical reason is that laser-based depth finders recover a depth measurement but not a color measurement. The "color" measurement that comes back from the laser system along with triangulated depth is usually a grayscale value of low quality. The depth measurement is in a coordinate system different from the one the camera uses to obtain the color measurement. The laser is usually swept rotationally, giving measurements in a rotational, or spherical, coordinate frame. The camera is a projective device and gives color samples in a projective pixel plane. In order to match up these two coordinate frames to associate the laser-based depth sample with the correct camera-based color pixel, a transformation must be computed. The transformation is usually computed by using a set of reference points as basis markers, and because this process is not perfect, the alignment suffers. It is not impossible to solve (there are laser scanners which recover high-quality color samples) but is a critical and difficult step when using the laser-based system. The second reason for not using lasers is that some lasers are powerful and can damage the objects they illuminate, not to mention the eyes of the operator. Curators naturally worry about the damaging effects of lasers and resist using laser-based technology on sensitive materials, even when one can show that a particular laser will not damage the manuscript or the user.
Instead of lasers, then, we opted to use structured light techniques based on an inexpensive, off-the-shelf, light projector as a controllable light-emitting device. A light projector produces a wide frustum of light (the "geometry" of the emitted beam) with a spectral mixture (the "spectrum" of the emitted beam). Controlling the pattern within the wide frustum is the "structured" part of the structured light process. Making the projector emit an easily detectable pattern is the goal -- the structure of the pattern is known in advance. A laser light source has a different structure, directed to a point rather than a frustum, and has a coherent spectrum, composed of a single wavelength rather than a mixture of wavelengths.
By using a light projector, we could directly convert the existing two-dimensional acquisition system into a three-dimensional acquisition system. The projector is coupled to the digitization setup and used to illuminate points on the surface of the manuscript page. The projector turns on its pixel P(x, y) and illuminates a three-dimensional point, M, on the surface of the page. The camera observes this illuminated point as a bright spot in its image at coordinate C(u, v). When the exact geometries of the camera and projector are known, this device-to-device correspondence can reconstruct the illuminated three-dimensional point, M. We recover the required geometries during a simple calibration step, completed before the digital scan commences. By repeating the projection and detection steps for each projector pixel P(x, y), the computer recovers a dense set of three-dimensional points from the surface of the manuscript leaf. We then use this dense set of three-dimensional points to create an accurate three-dimensional model of the scanned object.
First, we convert the set of points recovered in the scan to a structured representation in order to map and coherently display the texture of the digital image. This conversion is done by connecting the individual points in a triangulated mesh. The mesh captures the space in between each sample point and provides a large set of three-dimensional "faces" onto which the digital image is rendered. Other conversions and internal representations, such as height maps, are also used to facilitate a rapid and interactive display for the end-user. The result of the scan is a coherent representation of the surface of the manuscript. Seales and Brown have estimated the accuracy of the scanning process to be on the order of 0.5 millimeters variation in depth. At that level of accuracy they can capture very small variations in the surface of the vellum and make accurate measurements of the size and volume of the various features on the manuscript:
[2-D image of mesh goes here]
A two-dimensional rendering does not do justice to the three-dimensional modeling, which is better represented in video.24
This and related research into three-dimensional imaging promises to aid in the reconstruction of damaged manuscripts, in this case with a seamless rejoining of fragments 6 and 50. A process called "mosaicing" can stitch together digital images from different regions, even when the objects are digitized separately with potentially varying scales, to form a seamless global portrait. The same process of mosaicing will allow us to fuse images acquired at different times under different lighting conditions (for example, with bright light, ultraviolet, and fiber-optic backlighting) to achieve a more complete and legible result. Because the manuscript fragments are often quite distorted by heat, water, bending, or other factors, the computer scientists will adapt a process called "image warping," or digital stretching, to return the three-dimensional shape of the manuscript page to a planar object. We will, in other words, use the geometry of the three-dimensional image as the basis of a "post-warping" to undo the damage that caused the warping. These processes are of course ideal occasions for close collaboration between the computer scientists and the humanities scholar, because the latter must decipher the text in the damaged regions and assess the accuracy of the digital restoration.
In addition to developing methodologies for restoration, a chief aim of the project is to access and search damaged manuscripts using a new representation that supports fast and efficient search strategies over the images themselves and the structured information that is added to the image collection throughout the editorial process, such as text and related commentary. Under the direction of James Griffioen, graduate students in Computer Science at the University of Kentucky have designed an innovative multimedia database capable of storing a wide range of data formats, including images, transcripts, editions, and glossaries of damaged manuscripts. The key to the database is its ability to store, search, and modify large quantities of metadata, auxiliary information entered along with the data or, in the case of images, digitally extracted from the manuscript images. Because the amount of metadata can exceed the amount of original data by several orders of magnitude, it is crucial to be able to store and search the metadata efficiently and quickly.
Another important component of this work is the development of a process to identify image objects (frequently, in this case, severely distorted or fragmented letterforms) and store this extracted information as metadata for future content-based searches using fast select and join operations. The database model is also able to record correlations between different stored data elements even when the data have dissimilar formats -- for instance, a region of a manuscript image that corresponds to a word in the associated transcription. Once established, the correspondence between the data, as well as both the image and the text, are also stored in the database. With conventional approaches, the transcript of a manuscript might be stored in one file, the edited version in another file, and the associated images in other files, with no way to link or search the various pieces of the digital collection simultaneously. By storing all the data, along with correspondence identifiers in a single database, applications can easily establish connections between distinct data elements and quickly search for and retrieve related pieces of information.
Because the database was built using the Java Database interface (JDBC) and Java's Remote Method Invocation (RMI), any number of interface programs run on any computer in a network can access the system. In theory it does not matter to the database how users enter data or metadata in the database, whether the information comes from a document marked-up with TEI, XML, or SGML tags, a graphical tagging tool, a standard graphical database interface, CGI-bin scripts, an image processing routine, a three-dimensional scanning program, or any other means. The database is designed to store large amounts of data and metadata having different formats (here mainly text and images), to allow inserting, updating, and searching these data and metadata quickly, and to access everything from anywhere. With appropriate interfaces, users can share the stored data in different geographical locations, facilitating collaboration among physically distant editors. Furthermore, because JDBC is based on SQL, standard SQL queries can specify powerful searches. Applications that wish to access the data stored in the database issue SQL queries on Java objects to obtain result sets containing the desired information.
The one drawback is that SQL is a foreign language to humanities scholars. In one of his miracles lost to both Otho B. x and Vitellius D. xvii, Saint Basil helps one of his disciples who wants to learn Greek the easy way, without studying it (ll. 512-523). In lieu of a miracle the humanities team has begun to develop graphical user interfaces to the database to provide intuitive facilities for searching the image and textual information that will be stored in the database. One of the aims of this project is to provide comprehensive glossaries that will help students translate a text to develop interfaces for the humanities editors themselves, to facilitate primary editorial tasks, such as building glossaries based on the images and the texts they are attempting to restore.
The resulting glossaries are displayed in a network browser for convenient editing, and later for reading and browsing, but they must also be searchable in user-defined ways through the database. For editors we have developed a Glossary wordlist including each word and its location from a properly prepared transcript or edition. Next, the tool provides a comprehensive group of templates for each part of speech for editing the wordlist into a completely tagged glossary.
The Glossary Tool thus enables the editor to tag each lexical item for complex searches through the database. The Tool helps maintain the relationship between image and text, moreover, by assigning both folio-line and edition-line -- in this case, folio 5(30 recto line 3 and verse line 2 from the Life of St. Basil the Great -- for each word surviving in a manuscript fragment. After defining the word and filling in the remaining grammatical information, the editor clicks Save and the tool writes the metadata to a local file. When the editor clicks Export to XML/HTML, the tool forms a valid XML file from the glossed word data. The file is also marked up in HTML so it can be viewed in a web browser. The visual display capability of the exported file facilitates editing the transcript of the folio image:
<glentry>
<hdwrd>cildhad</hdwrd>
<pos type="noun"><noun gdr="m">m. </noun></pos>
<def>childhood, infancy</def>
<glform>
<nform decl="st" case="dat" num="s">ds. </nform>
<glwrd>cyldhade</glwrd>
<loc><foline>5(3)r3</foline><edline>2</edline></loc>
</glform>
</glentry>
All the entries, collated by headword, are simultaneously tagged for display, as
illustrated below.
After the editing process is complete, the editor creates a separate file that stores SQL insert commands for each glossed word entry, and imports the complete glossary data to the database for storage and searching. We are currently developing other interfaces to facilitate the editorial process and to provide access to users of the database.
The costs and difficulties of inaugurating a digital library of complex, image-based, electronic editions have to be balanced with the benefits of widely disseminating unique and therefore relatively inaccessible cultural documents while developing new technological capabilities. The collaboration of computer scientists and humanities scholars is still a highly unusual and therefore problematic situation in academia today, and it would require profound changes in university structures for this kind of interdisciplinary work to occur on a normal basis. Computer scientists tend to work on short-term projects with external grant support in Colleges of Engineering, while humanities scholars tend to work on long-term projects without regular grant support in Colleges of Arts and Sciences, Colleges of Liberal Arts, or Colleges of Fine Arts. We are learning by the frustrating process of trial and error, misunderstanding, and miscommunication, just how difficult this odd coupling is. To succeed in the long run, the newly emerging field of humanities computing requires a stable and dedicated source of system administration and programming support. Just as we require for our project interfaces to negotiate the SQL divide between the electronic editions and glossaries of the humanities scholars and the novel supporting tools of the computer scientists, university structures must provide administrative "interfaces" between colleges to encourage truly collaborative enterprises. Without an infrastructure for system administration and programming support for humanities computing projects, the short-term results of the computer scientists will quickly become unusable or obsolete in the long-term digital library projects of the humanities.25 It remains to be seen whether the Digital Atheneum project will offer new solutions to this underlying problem as it explores and develops new techniques for restoring, searching, and editing humanities collections.
|
Notes
1. Supported by the National Science Foundation's Digital Library program, IBM's Shared University Research grant, the British Library, and the University of Kentucky Center for Computational Sciences, the project combines the expertise of computer scientists and humanities scholars in an effort to make newly accessible some of the most badly damaged manuscripts from the Cottonian collection in the British Library. The principal investigators are Kevin Kiernan, Brent Seales, and James Griffioen, with the assistance of Linda Cantara, C.J. Yuan, Katherine Wenger, Michael Brown, Michael Rogers, Demorah Hayes, Kenneth Hawley, and Ashwin Gokhale at the University of Kentucky, and David French at the British Library.
2. The codex is in such desperately poor condition that the British Library has taken it out of circulation and the conservation laboratory is reviewing various extreme methods to halt its progressive deterioration. One problem is that the fragments were unfortunately framed in acidic paper in the nineteenth century, and the paper frames are both disintegrating around the manuscript leaves and staining them. The brittle vellum leaves have themselves crumbled in places, and previous conservators have sometimes used ill-advised methods to hold them together. One such method was gluing pieces of gauze over brittle vellum, which rendered these passages unreadable as the glue aged and became opaque.
3. For the standard introduction to St. Basil's life and works, see Paul J. Fedwick, "A Chronology of the Life and Works of Basil of Caesarea," Basil of Caesarea: Christian, Humanist, Ascetic, vol. 1, ed. Paul J. Fedwick, Pontifical Institute of Mediaeval Studies, Toronto, 1981, pp. 3-19.
4. Max Förster, Zur Geschichte des Reliquienkultus in Altengland, Sitzungsberichte der Bayerischen Akademie der Wissenschaften, Phil.-Hist. Abt., Jahrgang 1943, 8 (Munich), 63-80. See also Patrick Conner, Anglo-Saxon Exeter: A Tenth-Century Cultural History, Boydell Press, Woodbridge, Suffolk, UK, Rochester, NY, 1993.
5. Francis Wormald, ed., English Kalendars Before A.D. 1100, Henry Bradshaw Society 72, London, 1934. According to Michael Lapidge, "St. Basil is commemorated in a large number of Anglo-Saxon calendars - seven - but always on June 14. Four of the calendars in question are from Winchester. In commemorating St. Basil on January 1, therefore, Ælfric was not following Winchester use" (p. 123), "Ælfric's Sanctorale" 115-129, in Paul Szarmach, ed., Holy Men and Holy Women: Old English Prose Saints' Lives and Their Contexts, SUNY Press, Albany, 1996; but as Patrick Conner points out, two of these calendars also give the date as January 1, one of which comes from New Minster, Winchester (Cambridge, Corpus Christi College MS R. 15.32, page 15).
6. The manuscripts, with the exception of one alienated leaf, are all in the Cotton collection of the British Library: Cotton Julius E. vii, Cotton Vitellius D. xvii, and Cotton Otho B. x. A fragment of a Latin vita paleographically dated in the early tenth century was found in a binding in Exeter; see Conner (28-29).
7. Ælfric's Lives of Saints ... edited from British Museum Cott. MS. Julius E. vii with variants from other manuscripts, ed. Walter W. Skeat, Vol. 1, EETS OS 76 & 82. The contents and organization of Cotton Otho B. x is significantly different from the manuscript Skeat uses for this edition. The other manuscript is Cotton Vitellius D. xvii.
8. The Complete Corpus of Old English in Electronic Form, ed. Antonette di Paolo Healey, with Richard Venezky and Peter Mielke (Dictionary of Old English Project, Centre for Medieval Studies, University of Toronto, January 2000).
9. See Andrew Prescott, "'Their Present Miserable State of Cremation': the Restoration of the Cotton Library," Sir Robert Cotton as Collector: Essays on an Early Stuart Courtier and His Legacy, edited by C.J. Wright, British Library Publications, London, 1997, pp. 391-454.
10. Not all of the missing Cotton manuscripts were destroyed in the fire. Prescott has recently informed us that the wife of David Casley, deputy librarian of both the Royal and Cotton libraries, on at least one occasion gave a visitor to the library "two bundles of MSS" from the damaged Cottonian collection! (Prescott, note 52). The Bodleian leaf allows one to imagine the dismal scene outside Ashburnham house in the days following the disastrous fire.
11. Given this situation, perhaps a more virtually real e-foliation would leave the Bodleian, Oxford, leaf out of the numbering and name BL folio 50 "fol. 8(50)" and BL folio 6 "fol. 9(6)." The problems with this solution are that a reader will not know where the Bodleian leaf belongs, while at the British Library the sequence 8(50)+9(6) for the recto would be reversed on the verso as 9(6)+8(50).
12. Skeat was even suspicious of textual corruption. Observing that "there is an abrupt transition here," but pointing to line 633 (where the story of the sinful woman suddenly resumes), he concludes that "nothing is lost" (p. 83, n. 1).
13. The hermit's ineffectiveness probably reflects Ælfric's attitude that active, socially engaged monks were more admirable than solitary, contemplative hermits. See Mary Clayton, "Hermits and the Contemplative Life in Anglo-Saxon England," in Szarmach (147-175).
14. The first published experiment using ultraviolet with a digital camera occurred with this manuscript in 1993, at the start of the Electronic Beowulf project. See Kiernan, "Digital Preservation, Restoration, and Dissemination of Medieval Manuscripts," Scholarly Publishing on the Electronic Networks: Gateways, Gatekeepers, and Roles in the Information Omniverse, eds. Ann Okerson and Dru Mogge, Association of Research Libraries, Office of Scientific and Academic Publishing, Washington, D.C., 1994, pp. 37-43. The project won the 1994 Library Association / Mecklermedia Award for Innovation Through Information Technology for pioneering the use of fiber-optic backlighting in electronic editing. See also Kiernan, "Digital Image Processing and the Beowulf Manuscript," Literary and Linguistic Computing 6 (1991), pp. 20-27.
15. Although Wanley before the fire correctly transcribed gehaten (p. 191), Skeat says, "I read it haten, as noted on p. 50" (p. 544, note 1); and N.R. Ker in his Catalogue agrees with Skeat.
16. "Of this homily there are two other copies, viz. in MSS. 0. and V., both of which are much burnt. I give such variations as I could decipher" (p. 545).
17. See 8(RQe20r)v 11-12:249.
18. See http://www.image.ox.ac.uk/.
19. See Prescott, "'Their Present Miserable State of Cremation"'.
20. An apparently successful method of restoring the suppleness of vellum is discussed in I. K. Belaya, "Softening and Restoration of Parchment in Manuscripts and Bookbindings," and "Instructions for the Softening of Parchment Manuscripts and Bookbindings," Restaurator 1.1 (1969), pp. 20-48 and 49-51.
21. Vitellius D. xvii is also rendered more legible with magnification, UV, and image processing; the script makes it immediately evident, however, that the leaf from Otho B. x does not belong to it.
22. The spine has "Homilies for Saints' Days, Brit. Mus., Cotton Ms. Vitellius D XVII;" see Wanley in Hickes, ii, p. 206, for its description before the fire. The manuscript leaves are now much tinier than surviving fragments of Otho B. x, but they are more skillfully or carefully inlaid in non-acidic paper; even the unfortunate gauze reinforcement was well done, for the glue does not obscure the writing beneath, as it often does with Otho B. x. Many of the leaves of Vitellius D. xvii are out of order and reversed, however, suggesting that Madden had difficulty reading it too. St. Basil is on fols. 79v-83r2l.
23. A vivid example even in 2D is the image of fol. 5(3)r above, which clearly shows how the paper frame has buckled in reaction to the shifting shape of the supple vellum it is supposed to hold in place.
24. A video is available in Kiernan's PowerPoint presentation, "Creating Electronic Editions from Medieval Manuscripts," on the website of our project, "The Digital Atheneum: new techniques for restoring, accessing, and editing humanities collections." [http://www.digitalatheneum.org]
25. We would like to thank John Connolly, director of the Center for Computational Sciences at the University of Kentucky, for his leadership in providing system administration and programming support for the humanities PI.
|