The language code in the record as a whole is used for the 245; since 740s can be in different languages, different protocols are used.

Good point about language recognition on short strings.   My main point was that we should be thinking "outside the box" re non-filing indicators.   Reinhold Heuvelmann's post alludes to several possibilities.  He writes " While it may be true that there are technical provisions to try and figure out, I'd still prefer to rely on the people cataloging, and their experiences."   Our experience with the LTI fix indicates that while catalogers as a group have a lot of knowledge about the ins and outs of non-filing indicators, there is a high error rate even at an institution that has traditionally prided itself on examining copy carefully.   The ideal system should use programming based on human experiences to  identify human error, with fine tuning for odd cases that look like error.

Amy

-----Original Message-----
From: Bibliographic Framework Transition Initiative Forum [mailto:[log in to unmask]] On Behalf Of Tom Emerson
Sent: Monday, February 11, 2013 9:22 AM
To: [log in to unmask]
Subject: Re: [BIBFRAME] Filing indicators

Amy Turner writes:
[...]
> They work from the language code, lists of articles by language, lists 
> of exceptional phrases "A is for ...", and specific cases of titles 
> that are not in the language of the text.

Do you find that records have valid language codes for specific fields like titles? In my experience the language codes, if they exist, can be inconsistent.

> Language recognition software, rather than the language code, could be 
> used (Following LTI processing, we corrected hundreds of language 
> codes that erroneous non-filing fix had revealed).

On short strings language recognition is very difficult: I haven't had much luck with it on strings under 250 characters or so.

    -tree