Report investigating automatic metadata generation tools

Just found this JISC final report: MetaTools - Investigating Metadata Generation Tools [pdf]. If you are interested in the automatic generation of metadata and current tools out there for generating Dublin Core metadata, this report is for you.

(It also covers the development of SOAP and RESTful Web Service interfaces for three metadata generation tools.)

This report is very interesting for me (which is why I just printed off all 98 pages!) because this is the direction we are trying to move in for our digital library development at work. We're not there yet, but "iterative development" is the mantra.

It does mean letting go of creating metadata records one at a time as a final product. This is the shift we all keep hearing that catalogers need to make in order to become metadata librarians. Having been a metadata librarian for a year now, I can honestly say it's hard to let go of the old mindset at first. (And depending on your metadata workflow you may not have to.) But I've had to and have survived the process. In fact, I think if I went back to AACR2/MARC cataloging I'd be a better cataloger. Just a thought.

Comments

[Laurel Tarulli]

Thanks for this Chris! I know what I'll be reading this weekend. With the emphasis on streamlining cataloguing practices and additional responsibilities cataloguers are taking on, I'm very interested in this. I, too, have a hard time with the idea of moving away from the control over creating/editing each record. Although, there are some great tools (such as MarcEdit) out there for mass edits. With the growth of technology and shifts in our own roles, I think we should all be reading reports such as this.

[Matt Ostercamp]

I'm curious. What do you think you've learned as a metadata librarian that would make you a better MARC cataloger?

[Chris Schwartz]

Hi Laurel, Matt,

Thanks for the comments. Matt, I thought about this and here's what I came up with:

1) In general, I'd probably worry more about accurate MARC coding.

2) I use to look at MARC records in terms of migration to another ILS. Now, I would look at MARC records in terms of crosswalking to other metadata schemas.

3) I would be more careful that important identifiers were not buried in notes (knowing that things in notes may be difficult to machine parse later).

4) I would make sure, if possible, that every MARC record had an 856 field linking to: full-text, an abstract, a book review--any digital version of an item or metadata about that item on the Web.

5) I'd worry less about pre-coordination of LC subject headings and more about getting more subject headings into the MARC record. My reasoning: to provide good keywords and bring out hidden aspects of an item/work.