When cm met CD, or, Why cataloguers need to care about how they talk to machines / Bernadette O'Reilly
•Download as PPTX, PDF•
0 likes•111 views
Lightning talk paper presented at the Cataloguing & Indexing Group conference in Swansea, UK in 2016.
Recommended
More Related Content
More from CILIP MDG
More from CILIP MDG (20)
Recently uploaded
Recently uploaded (20)
When cm met CD, or, Why cataloguers need to care about how they talk to machines / Bernadette O'Reilly
- 1. When cm met CD, or, Why cataloguers need to care about how they talk to machines Bernadette O’Reilly, Catalogue Support Librarian, The Bodleian Libraries. 31 August 2016 Images: ©2014 ClipartPanda.com
- 2. • Nearly 200 cataloguers • 100 institutions • Many not employed by Bodleian
- 3. When cm met CD, or, Why cataloguers need to care about how they talk to machines Bernadette O’Reilly, Catalogue Support Librarian, The Bodleian Libraries. 31 August 2016 Images: ©2014 ClipartPanda.com
- 4. Travel documents ... Male, fair, tall, young Female, dark, short, less young Jail image: ©2014 ClipartPanda.com
- 5. ... for metadata LDR cam a2200385Mi 4500 005 20150804045511.0 008 150428t20152015fr g b 001 0dfre d 016 7 |a FRBNF443319120000005 |2 FrPBN 020 |a 9782200601508 (br) 020 |a 2200601506 (br) 024 30 |a 9782200601508 035 |a (OCoLC)908129047 040 |a ZWZ |b fre |e ncafnor |c ZWZ |d BDF 082 04 |a 306.8 1001 |a Kaufmann, Jean-Claude |d (1948-....). |4 aut 245 10 |a Casseroles, amour et crises : |b ce que cuisiner veut dire / |c Jean-Claude Kaufmann. 250 |a [Nouvelle édition]. 260 |a Paris : |b A. Colin, |c DL 2015, cop. 2015. 300 |a 1 vol. (366 p.) : |b couv. ill. ; |c 22 cm. 336 |b txt |2 rdacontent 337 |b n |2 rdamedia 337 |b n |2 isbdmedia 504 |a Bibliogr. p. 347-358. Index. 650 7 |a Cuisine |x Aspect social |z France |y 1990-.... |x Enquêtes. |2 ram 650 7 |a Alimentation |x Aspect social |z France |y 1990-.... |x Enquêtes. |2 ram 650 7 |a Famille |x Sociologie |z France |y 1990-.... |2 ram 650 7 |a Sociologie du quotidien |z France |y 1990-.... |2 ram 85642 |3 Notice et cote du catalogue de la Bibliothèque nationale de France |u http://catalogue.bnf.fr/ark:/12148/cb44331912w
- 6. Shock I: Foreign-language records
- 8. Strategies • Global correction and downgrade for miscoded records. • System fix for newly-added records: – Code as new – Remove 035, 015, 016, etc. – Assign new 040 field with $b = eng – Warning field. • Documentation • Automated warnings for records with non-English 040 $b. • Frequent reminders. • Individual followup.
- 9. Not quite there yet 015 __ |a 10,B19 |2 dnb 016 7_ |a 1001983610 |2 DE-101 035 __ |a (OCoLC)614493721 040 __ |c GWDNB |d HEBIS |d OCLCQ |d UkOxU 082 04 |a 320.9624 |2 22/ger 084 __ |a 320 |2 sdnb 24500 |a [Title redacted] / |c ed. by the Heinrich Böll Foundation. [Alex de Waal ... Ed.: Robert Furlong ; Bernd Herrmann]. 260 __ |a Berlin : |b Heinrich-Böll-Stiftung, |c 2010. 300 __ |a 123 S. : |b ill., Kt. ; |c 24 cm.
- 10. Why does it matter? • English-language catalogues should be English-language. • Different cataloguing rules, name authorities, LCSH - may impair searching. • Records may be rejected by OCLC. • If records are not rejected by OCLC, other agencies will get unsuitable records.
- 11. Shock II: ‘Unresolved’ records • Many records – had an OCLC system number in 035; but – had been edited to be used for a different resource. • If our records are rejected, WorldCat users will not know that we have the resources. • If our records slip through and cluster with the wrong master records, WorldCat users will think that we have copies of resources which we do not have.
- 12. ‘Same’ record – different resources 035 |a (OCoLC)19508660 1001 |a Varennes, Claude de. 24513 |a Le voyage de France, dressé pour l'instruction & commodité tant des françois que des estrangers. 250 |a 4. ed., corr. & augm. 260 |a Roven, |b J. Caillové, |c 1647. 300 |a 10 preliminary leaves, 60, 304, [29] pages 035 |a (OCoLC)19508660 1001 |a Varennes, Claude de. 24513 |a Le voyage de France, dressé pour l'instruction & commodité des François & Estrangers. : |b Avec une description des Chemins, pour aller & venir par tout le Monde. Tres-necessaire aux Voyageurs. / |c Corrigé & augmenté par le Sieur Du Verdier, Historiographe du Roy.. 260 |a A Paris, : |b Chez Michel Bobin, au troisiesme Pillier de la Grand'Salle du Palais, à l'Esperance., |c 1655.. 300 |a 8 unnumbered pages, folded map, 492, 4 unnumbered pages : |b illustrations; |c 8°.
- 13. Innovation means ... ... caring and sharing ... with the power of machines. Image by VectorCharacters.net
- 14. css@bodleian.ox.ac.uk Bernadette O’Reilly Catalogue Support Librarian The Bodleian Libraries.
Editor's Notes
- Hello. I’m Bernadette O’Reilly, and I work for the Bodleian Libraries in Oxford. My job is to keep up with developments in international cataloguing standards and provide training, documentation, advice and a bit of quality control ...
- ... not just for the Bodleian’s own cataloguers but for all the cataloguers in Oxford University’s Libraries Information System (OLIS). So I actually work with [click] nearly 200 cataloguers [click] in nearly 100 libraries scattered over several miles, half of which (the libraries in the colleges and some departments) [click] are completely independent of the Bodleian. I can’t be looking over everyone’s shoulders – I actually work a mile and a half from this pretty bit, in the industrial estate beyond the top right-hand corner - so quality control depends on reports, when we know what to look out for, and otherwise on serendipity – just happening to notice records which indicate misunderstandings or knowledge gaps.
- So sometimes I get surprises. All the big surprises in the last few years, the drop-everything-to-fight-this-fire surprises, have involved the kind of data which is used for managing records both internally and externally – the elements which control things like whether a record should be edited, or upgraded, or overwritten, or exported to COPAC or WorldCat and, particularly, how our exported records will be handled by the receiving agency. Fires of this kind have proved particularly difficult to fight – smouldering unnoticed for years, taking hold in several locations at once and with a trick of breaking out afresh when we hope they have been damped down. I would like to share what we have learned from a couple of these surprises, both of which relate to importing and exporting records. We know that the problems are not confined to the Bodleian. Some of our faulty records were actually copied from other databases with the problems already present – from OCLC, RLUK and even LC and BNB. The problems indicate that our cataloguing culture has not kept up with our the changing cataloguing environment. It is very clear that our cataloguers find it much easier to appreciate the importance of good descriptive data and access points than to appreciate the importance of the kinds of codes and identifiers which are primarily intended to be read by machines for automated processes. But this data is absolutely key to our ability to meet the growing challenges of dwindling resources and the growing opportunities of linked, machine-actionable data. I am not at all intending to criticise my colleagues (or anyone else’s). When I was first trained to catalogue just 11 years ago I did not need to know anything much beyond AACR2, MARC21, LCSH and our local software. Now our cataloguers are expected to make best use of a vast and varied network of data and processes, both local and external. It’s a big ask. It is not enough just to design and document ‘efficient’ new streamlined workflows and algorithms – they will not work efficiently until we win the hearts and minds of all our colleagues to care about all the kinds of data which the workflows and algorithms use.
- You could say that these issues are about passport control [click for passport control image]. Off to Paris next week, but whoops! I forgot to renew my passport. No problem – I’ll borrow my son’s. Of course, he is [click] male, fair-haired, six feet tall and half my age, while I am .. different [click], so I’ll have to make a few adjustments. What could possibly go wrong? [Click for jail image]. Well, I might not get to travel at all. If the Border Agency is cutting corners that day, I might get through. But that could cause another set of problems. Records might then indicate that my son is in France, while I am still in the UK. If someone really needed to get hold of one of us, they might get the wrong person, or no one at all. Because, although I changed the descriptive bits on the passport, it still has a serial number which is associated uniquely with my son, and so goes with his description, not mine.
- Nowadays catalogue records are travellers too. Many agencies continually import records from WorldCat, RLUK, the Library of Congress, the British Library and others, and export their own records back to WorldCat and RLUK. And records, like human travellers, have identities and histories, which are enshrined in various serial numbers and other identifiers. These associate them with a particular resource, and by implication with key aspects of its description, e.g. title, content, date of issue, physical format. For instance, this record from OCLC’s WorldCat has - [click] in the 035 field, its OCLC system number; this will be preserved if the record is downloaded to another database - [click] in the 016, a national bibliography number; other records might have a BNB number in 015 and/or a Library of Congress control number in 050 - [click] in 040, codes of the agencies which created or edited the record, giving it a history. Nowadays records even have something like nationalities, the language-of-cataloguing-agency recorded in 040 $b[click] , in this case French. This, of course, is not the language of the resource, but the language used in the non-transcribed elements of the record, e.g. for physical description [click] and notes. WorldCat now has records from many different language communities, and, to allow users to find records in their own language, has ruled that a different language-of-cataloguing means a different record. That means that the identifying numbers in a record’s 035 fields are associated not just with the resource catalogued but also with the language in which it is catalogued. And, of course, national bibliography numbers are always associated with a particular language.
- Shock the first: This one hit us in the summer of 2012, just when we really did not want to have to think about anything except RDA implementation. A record got itself noticed because it did not use Unicode and so displayed oddly in our system [click]. Its 040 $b and $e showed that it was created by a German-language agency to German standards [click] (Regelwerk fur Alphabetische Katalogisierung Wissenschaftliche Bibliotheken), and in other respects it was quite unsuitable for our use – physical description in German, no English-language subject headings – but that was not the shock. Initially we thought that it was more or less a one-off, and all we needed to do was to make sure that everyone realised that OCLC included non-English records, and knew what to look out for and avoid. The shock was when we were told by several managers that their cataloguers could not manage without downloading non-English records, and had been downloading them for years. (Particularly cataloguers who dealt with foreign-language resources and art resources.) In fact, there were already several thousand records in OLIS coded as non-English. We just hadn‘t noticed before. Many had been more or less fully translated and adapted to English-language standards, but were still coded as French, German, Dutch, Spanish, Italian....
- ... while a few had not been translated or adapted at all. And that is how CM met CD in our catalogue [click] – ‘met’ is Dutch for ‘with’, so the CD is just accompanying material. (I whited out the title to avoid embarrassing any individual cataloguers.) Our cataloguers had not been specifically warned to look out for foreign-language records, and many of them were unaware of the significance of the code in 040 $b. It had never occurred to me or my immediate colleagues that anyone might be tempted to use a record which had funny 300 fields and used unfamiliar standards and authorities and subject headings, rather than cataloguing originally. That was a failure on my own part to engage with this new source of data, because when original cataloguing would involve lots of diacritics or nonroman scripts and transliteration, adapting even a foreign-language record starts to look good.
- So we had to get ourselves sorted fast. [click] For the thousands of existing OLIS records with non-English 040 $b, we could only do a rough global correction, because there was so much variety as to how far they had been checked and/or translated; so all those records had to be downgraded. [click] A brilliant systems colleague designed a fix to make it possible to use downloaded foreign-language records safely in future. It recodes the records correctly as new English-language records, and adds a warning field which reminds cataloguers that they still have to check MARC, RDA, LCSH, etc. themselves. The field should not be deleted until this is done, and meanwhile it prevents export. Problem solved? Actually, it has taken a stupendous effort to get through to (almost) all our cataloguers the importance of using the fix and making the checks: not just documentation [click], automated warnings [click] and general reminders [click] but in many cases person-to-person [click]. Communicating, providing context, mentioning the problems which might arise, so that it stops looking like a technicality. (This was obviously more difficult for us because our cataloguers are so scattered.) Even very professional people could find it really hard to grasp that changing the language code and clearing out all the record’s old identifiers was as important as translating the data. After four years things are much better, but still a few people assume that they can make the problems go away by making little edits to the elements which trigger warning messages rather than using the proper fix and procedures.
- In this record the cataloguer deleted the 040 $a and $b, thus losing the ‘ger’ code, to make the warnings go away; but it still has the 035 field which identifies it with a German-language record in WorldCat. (In this case the physical description [click] is still in German, but that kind of error is much less common.)
- Does it really matter? - It’s obvious that people should be careful about what they download and translate it if necessary. Users expect English-language catalogues to be in English, so the records ought to be thoroughly translated. - And it’s not just a matter of translating. Foreign-language records often use different cataloguing rules, different versions of MARC, different name authorities and different subject systems, so all of that ought to be checked as well, to maintain effective indexing. - But there are potentially much greater problems if records are translated and are then exported again, particularly to WorldCat, without being given both the correct language-of-cataloguing code and a new identity. If they still have OCLC’s original system numbers in 035, OCLC’s deduplication algorithm will expect to cluster them with the original master records. But the language of cataloguing code, if it has been changed, or the translated terms in the physical description, etc., will preclude matching, so the records will probably be rejected as anomalous. In consequence, people searching WorldCat will not know that we have copies of the resources. - If some records with unchanged language codes do slip into OCLC, they will be a snare and a delusion. Perhaps a Spanish-language agency wants to harvest Spanish-language records, so selects by language of cataloguing, 040$b=spa; but they might well end up with a proportion of records which have actually been translated into English. Rather like if I try to travel on my son’s passport [click]: either I don’t get to travel at all, or I slip through the checks and then people who want to find my son get me instead.
- Shock the second After our last system migration it took a while to get our export to OCLC going again. When we eventually exported the backlog, several thousand records were rejected. Many for the language-related reasons I have just been talking about, some because of technicalities such as nonroman script in the wrong fields; but very many simply because of other types of mismatch Lots of the records [click] had been derived from OCLC and still had the original system number but had been edited in ways that made them seem to be for different resources. OCLC’s matching algorithm could not determine whether the records should be assimilated to existing master records or treated as new, and so rejected them entirely. To be fair, sometimes they really were for the same resource but had been improved out of recognition, e.g. if a brief record had been edited to antiquarian standard – and in that case the only solution is to get the OCLC record improved; but most often they really had been edited to be used for a different resource – a different edition, a different date, a microform record used for a print resource, a single-part record extended for a multipart resource… Again, what was startling was that this was being done not just occasionally by oversight, but quite regularly by experienced and careful cataloguers. One OCLC record had been downloaded 4 times, and adapted for 4 separate editions of a work, so we had 4 different OLIS records all with the same OCLC number in the 035 field.
- An example: here are the key fields of an OCLC record; and here [click] are the key fields in an OLIS record derived from it. The OCLC number in 035 is retained in the OLIS record, but the edition statement, place, publisher, date and physical description are all completely different, showing that it is being used for a quite different resource, although in other respects it is a very good record. In fact, it was trying to travel on another record’s passport [click], and was caught. This particular hearts –and-minds campaign has only been running for 9 months, and most people have got the point now, but still people sometimes forget.
- Basically, this is all about sharing and caring, but [click] on a scale so large that it can only be managed by mechanised processes [click]. By exchanging data on a really large scale we keep our operations viable, enrich our services and support other information providers. We have to be able to select and manage data on the basis of machine-readable markers; and so we have to be able to trust each other to apply these markers accurately. Which means that cataloguers have to care about how they talk to machines.
- Have you encountered similar problems in your own libraries? Is there a future for encoding levels or other indicators of completeness and finality? Does anyone know of plans for record management data in Bibframe or similar?