TRANSLATE ENGLISH to Français, Deutsch, Italiano, Português, Español! Explanation
© LIBER and author
In April the British Library signed a contract to convert the published catalogues of printed and manuscript maps. These run to 19 volumes: the 15 maroon volumes covering printed maps up to 1964 (with its 10-year supplement), and the three mid-19th century brown volumes describing manuscript maps. Optical character recognition was never a serious possibility given the varied and unclear type and the hand-written annotations. We therefore went for keyboarding.
A detailed specification was drawn up, by James Elliott and then Geoff Hutt, to guide the keyboarding bureau. This includes instructions about the handling of headings, the capture of scripts, punctuation &c and, most importantly, the way to apply the correct field identification for conversion to UKMARC. The specification is about 100 pages long. To avoid getting lost in detail, much of it relating to the peculiarities of our own catalogue, I will not say more about it.
After a long tendering process (roughly a year) the contract to publish the resulting file on CD-ROM was awarded to Research Publications International, who themselves contracted the keyboarding to Access Innovations in New Mexico. The commercial principle underlying the arrangement is relatively simple. RPI receives exclusive publication rights to the BL cartographic file for 15 years. In return, the Library enjoys cheaper keyboarding and a share of the profits. We are still in the test phase so I can say little about how the project works in practice. Printout will be supplied to the Map Library so that the four-person team can proof-read samples. There must be no more than 1.5 errors per thousand for the most important elements and 3 per thousand for everything else. We plan to proceed at the rate of one volume per month.
After completion of the keying (scheduled for Spring 1994), concordances will be run against the files to impose consistency (e.g. on the precise wording of pressmarks) and to add further searchable elements. Once the printed and manuscript map files have been merged, and joined to the existing automated file for material described since 1975, RPI will issue a CD-ROM containing almost 250,000 records. This should be in the second half of 1994.
I assume that map librarians need no conversion to the idea of an automated catalogue. I wish to concentrate instead on the choices facing those who consider converting an old map catalogue. Much of what I am going to say has to do with 'quality', and you may be surprised to hear me refer to the search for quality as a problem rather than an ideal. I believe that this apparently worthy aim can in fact prove to be an important psychological barrier. The rest of my talk will therefore expand on the old saying that 'the best is the enemy of the good', that those who seek perfection may well end up achieving nothing.
The observant will already have realised that the BL's cartographic file is to be a timeless one: there is no date cut-off. In most libraries, automation has been applied only to current cataloguing. When conversion has taken place, as with the BL's general catalogue, the historical file is kept separate from the current. Any cut-off date is arbitrary and unhelpful, splitting editions of the same work. As far as the literature on the history of cartography is concerned, the records even end up in the wrong file. We therefore decided to create and maintain a single file. Treating mapping as a continuum from the earliest times to the present is logical. But it immediately brings you face to face with 'quality', because the catalogue descriptions also represent a long date-span. Inevitably, this means records of different style, completeness and accuracy.
When it was suggested, about five years ago, that we convert our catalogues of pre-1975 maps, my first reaction (I am ashamed to say) was to reject it outright. How could we sacrifice the quality of the current records by mixing them up with the old? The BL has an unusually large and broad collection but I hope that our conclusions may be relevant to any map collection with a sizeable proportion of earlier material. As we shall be the first, I believe, of the world's great libraries to publish our historical map descriptions in automated form, our decisions may well have bearing on international moves to exchange such records in the future.
There are two possible approaches to Retroconversion: the theoretical and the possible. The British Library has committed itself very firmly to the possible, even when this has meant accepting some degree of what librarians hate, inconsistency. Although most of our approximately 2 million maps are 20th century (perhaps 80%), only about 3% of the records refer to material published since 1975. This is because we do not describe series maps individually. Any long-established library that catalogues to series level will be in the same situation. For us, then, Retroconversion involves merging a very small number of recent records (created with machine-readability in mind), and a very large number of older descriptions (designed for traditional access).
What should we do with old records that we now consider 'inadequate' - like 'quality' another emotive word? One answer might be Retrospective Recataloguing. It does not take long to demonstrate its impossibility in a large collection like ours. Even with extra staff (a dream!), it could still take 50 or more years to recatalogue those items that were described before the 'bible' became available - in our case, the Anglo-American Cataloguing Rules (2nd Edition). For us, Retrospective Recataloguing - the 'best' - would mean the death of Retroconversion - the 'good'.
Perhaps we should look more closely at the supposed inadequacy of earlier records. Four types of deficiency can be identified in our catalogues: omitted information, inaccuracy, data expressed in the wrong way, and structural problems. In essence, our Retroconversion programme involves overcoming these as best we can. I will take the elements in turn.
Firstly omitted details. Since we do not have the time to re-examine maps, omitted information is likely to stay that way. Early cataloguing rules of the British Museum (as we then were) omitted both publisher and scale. So records created up to about 1940 may lack either or both of those elements. In these respects the converted catalogue will be no worse than its printed predecessor. Nor is scale of so great importance when dealing with early, non-series printed maps, whose scale can often be roughly deduced from the map size. Manuscript maps, on the other hand, have unpredictable scales. Happily, we can translate, by means of a concordance, the transcribed verbal scale statement into the Representative Fraction. A number of you may immediately know the RF for "15 Calenberger Ruthen to the inch"; others, like me, will be grateful for the help.
Further missing elements that we shall be able to supply by concordance are a country code, the normalised place of publication, a link between the record for a series and that for its index map, and the metric sizes of manuscript maps. We are also adding beforehand a thematic element (canals, military &c) to those entries where it is lacking. I will deal later with geographical access and absent dates, but those represent the main ways we shall overcome the lack of information in the original record.
Inaccuracy in the old descriptions is the second of the deficiencies I mentioned. When it involves authority-controlled data (geographical or author names) this will be put right when the headings are edited as a whole. Otherwise, and particularly where cartobibliographical knowledge has since advanced, it will not be possible to do anything at this stage. However, corrections can later be made whenever they are noted.
The third category concerned data expressed in the 'wrong' way. This typically involves differing styles of title transcription and has little relevance to computer searching. Finally, what I termed 'structural problems'. By these, I mean information in the wrong place, typographical inconsistency &c. These instances are noted in the Specification, whose rules for data capture should ensure that the converted file is, in these respects, consistent.
So, despite using the old descriptions and without looking at the original items, the British Library catalogue on CD-ROM will represent significant improvements to the content of the earlier entries - even before the enormous benefits of interactive access are considered. How much should we apologise for the inconsistencies that will be repeated on the disc? Very little, I feel.
It is essential to remember the distinction between a bibliography and a library catalogue. We expect from a cartobibliography to be able to distinguish similar maps, and we look for a full and clear statement of the bibliographical relationship of one variant to another. A library catalogue, on the other hand, should be judged firstly by how well it provides access to the geographical content of the listed material. The British Library Map Library now makes a clear distinction between bibliographical and topographical value. Minor sacrifices of bibliographical detail can speed up the cataloguing process, thus making more maps available. We see it as our task to lead the user, quickly and helpfully, to anything that might be of relevance. Thereafter, it is up to them to examine the items for themselves.
It is appropriate that the keyboarding agency involved in our Retroconversion is called Access Innovations because the fundamental purpose of the exercise is to make the Library's cartographic collections more readily and more fully accessible. This means exploiting the interactive possibilities to the full. We shall therefore be concentrating on headings and indexed elements, rather than unsearchable factors such as capitalisation in the title or the arrangement of notes fields. Uniformity of pressmark citation will be essential for the automated retrieval system we shall be using in a few years' time in our new library building at St Pancras. Otherwise, staff effort will be devoted to tidying up the incomplete matches made by the various concordances, and expanding and editing the author and geographical name headings.
It is presumably our common experience as map librarians that most map enquiries are expressed geographically. Geographical heading, and in particular geographical hierarchy, has played an important part in our planning. The major defect of our 15-volume catalogue of printed maps is its 19th century dictionary arrangement. To find general, regional and local material relating to Spain it would be necessary to predict every possible sub-heading or, more realistically, read through the 15,000 columns. This difficulty will be overcome on the CD-ROM. Having first made sure that all entries carry an 'address' - the name of the superior region or country - we will then apply by concordance a hierarchical code extracted from the 19th edition of the Dewey area tables. By this means it will be possible for a searcher to request, for example, all 16th century material relating to the whole of Spain or any part of it. Besides the British Isles, where the coding is taken as far as the county, the hierarchy descends no further than country level (except for USA and Australia). There is no reason why in the future the coding should not be refined further. Care needs to be taken, however, that the codes do not become more precise than the geographical knowledge of users.
Publication of the CD-ROM in 1994 will be an important event in map librarianship, but it will certainly not be the end of the project. This should be seen rather as a 15-year programme, involving both the enhancement of records already in the file and the addition of large numbers of historical records. The existing author headings, in what was essentially a geographical catalogue of printed maps, represent only a minority of the surveyors, cartographers, engravers and publishers mentioned in the entries. Extracting these hidden names, editing them and adding them to the name index on the disc will ensure the BL file becomes a major source for future work on mapmakers.
The long-term aim is a single overview of one of the world's great map collections, the creation of a comprehensive file containing all material of interest to both current map users and cartographic historians. I mention briefly some of the elements that will be added in the later phases: the map collection of the Royal United Services Institution; a large collection of War Office compilation material, much of it dealing with the first mapping of 19th century Africa; the Crace Collection of London maps; maps in books in the main library; maps in the Oriental and India Office Collections; documents of cartographic interest, such as letters from mapmakers, references to surveys, publishers' invoices &c; and the large number of manuscript maps acquired since the catalogue was compiled in the 1840s, many of them not described anywhere.
Part of the problem is that none of the elements I have just listed are treated as part of the cataloguing 'backlog' that dominates most librarians' lives. The same is true for the contents of atlases. In the case of text volumes this makes sense, but maps bound into atlases form the greater part of any historical collection. The BL has never catalogued atlas maps in a systematic way, nor have we have the staff to do so now. We were therefore delighted when Rodney Shirley, the well known cartobibliographer, volunteered to describe the contents of our pre-1800 atlases. These will be published in the form of collations, and the entries will also be added to a later edition of the CD-ROM. The records have not been created by a librarian and they do not fully conform to the complex AACR2 cataloguing rules. What matters is that we shall be able, for the first time in our history, to provide a full answer to questions such as: 'how many pre-1800 maps of Catalonia are there in the British Library?' The 'good' triumphing once more over the 'best'.
Successive versions of the published disc will not only add recent accessions, therefore, but also bring to light hidden aspects of the Library's collections. At least one map has remained uncatalogued for 250 years. It seems unarguable to me that it is more important to have some kind of record for every map than a perfect record for some of them. This does not rule out future improvement. Retroconversion should not be seen as a 'once and for all' operation, like publishing a map catalogue in book form. Records can be corrected or expanded at any time. The system can itself select sub-standard records, or identify a convenient group for recataloguing. The most serious defect of some of our own earlier records is the lack of a date. Since date will probably be used to refine most searches, this means that the records concerned would simply not appear. This is perhaps the most urgent of the future editing tasks.
Looking in one direction, the catalogue conversion project is a tribute to the many, mostly unknown, librarians who created the descriptions over a period of one and a half centuries. Looking towards the future, we can anticipate that the catalogue will be endlessly improved - often with the help of our expert users - or even reshaped to meet the needs of later generations. This plastic quality, this adaptability, fits well with our belief in the practical, not theoretical approach.
In preferring compromise to impossible perfection I must stress that we are not behaving irresponsibly. The records will be in a (simplified) UKMARC format and we very much hope that they can later be used as part of a European or global cooperative map file. Where the information is available it will be stored in the correct field, but it has to be understood that some records will be incomplete. If a union map catalogue is ever to be achieved, these limitations must be accepted. I strongly urge any who contemplate Retroconversion to focus on the benefits for users rather than the supposed deficiencies of the records.
Back to GdC Homepage