We've recently come across a problem to do with non-roman scripts (which we add in 880 fields) and authority records.  We use Millennium, but that's only partly relevant as there are some theoretical issues I would like some advice on.

Our general practice has been to create 880 fields in Chinese script for all the descriptive fields, where appropriate, and also for name subject headings and name entries for authors etc.  You can see an example here:
http://library.dur.ac.uk/record=b1774771~S1
(there is a MARC record button allowing you to see the MARC data).

Until recently we had been creating authority records manually, and not bothering with non-roman scripts in them, but now we are pulling full authority records from OCLC.  In the above example, the authority record for Kangxi, Emperor of China, has many Chinese script versions of the name as well as a multitude of transliterations and variants.

We run a report in Millennium daily which tells us about new headings added to the catalogue.  The Chinese script version of the name which we use in the 880 now comes up as an invalid heading, because it appears as a 400 on the authority file.  Similarly, if you check the headings list when creating an 880, the system will unhelpfully direct you to prefer the roman script version of the name.  A further issue is that anyone browsing catalogue by author under the Chinese version of the name used in the 880 will see the records but will also be referred to Kangxi as the preferred heading.  Before we had Chinese headings in the authority record, putting in 880s for the headings on the bib records meant the terms were treated equally in browsing, with no enforced preference for a particular script, which was good.

Setting aside how our particular LMS operates, how should this sort of cataloguing be done, in theory?

Should a bib record have a 880s in original script corresponding to X00 fields?

Advantages:
-- the Chinese version appears in parallel with the western version in the catalogue display
-- the user can browse Chinese headings without being told constantly to prefer a particular transliteration
-- on Millennium, if the Chinese heading was not in the bib and only on the authority record it would not keyword-index

Disadvantages:
-- if we keep the authority records as they are, we have invalid heading reports, and a redundant cross-reference when browsing headings
-- it relies on the cataloguer creating the 880 for the X00 fields consistently as Millennium does not automate this or ensure they are present, nor can the authority record tell us which Chinese script heading is preferred.

The advantages are all on the side of the user.  The disadvantages are only for staff.

We would prefer not to meddle with the authority records we bring in, as that defeats the point of downloading them, but one option to avoid the invalid headings and cross-references might be to pull in the authority record and delete all the Chinese script headings, but then we wouldn't get cross-referencing from alternative versions in Chinese script either.  If we pulled in the authority record a second time and this time deleted all the roman script headings, putting the preferred Chinese script heading in 100 and the variants as 400, that would work technically in Millennium but having two authority records for one person (one for each script) seems onerous and unsound as there should surely only be one authority record per person.

It seems to me that the authority records we get from LC do not convey sufficient information for the system to Do The Right Thing.  The Right Thing, to my mind, would be for there to be a preferred Chinese script heading, and a preferred English transliteration of the heading (and possibly a preferred French transliteration, German transliteration, etc.).  There would be added headings (400 fields) in the authority, which might have transliterated equivalents in 880 or might not.

With such a structure, the system could pull in the 100 field and the 880 fields for a person in tandem, and could display the Chinese equivalent of the preferred heading when browsing the Chinese names without forcing the user to follow a "see reference".

It seems from this web page: http://www.loc.gov/marc/authority/ecadmulti.html that it is allowable to use 880 fields in the authority records to provide this kind of structure, but does anyone do that in reality, and does your library system actually support it?  Even using 880 fields in this way it still does not help distinguish different styles of transliteration (e.g. Tchaikovsky, Tschaikowski, etc.) which are preferred by speakers of different languages.

We'd be very glad to know how other libraries cataloguing in CJK or Arabic handle this: does your system do things any better?  Do you just leave your 880 headings unchecked and essentially uncontrolled?

Thanks,

Matthew

-- 
Matthew Phillips
Head of Digital and Bibliographic Services,
Durham University Library, Stockton Road, Durham, DH1 3LY
+44 (0)191 334 2941