Auto-generating library catalog tags

February 14, 2017 9:29 AM

Auto-generating library catalog tags
I don't have much of a background in programming, but I do have a bit of one in libraries. This was my final paper for a graduate course from last semester. I'm curious to hear from other people in the library and programming fields about the viability of the idea. Is it useful, feasible, doable? The basic idea is to automatically attach search terms to library items that end up being used because of those search terms. I'd love to hear what you think!

Role: Creator

posted by holmesian (1 comment total)

I think this is an interesting idea (and sorry for lateness, we didn't cruise through Projects for the podcast this month). A few thoughts.

- anything that requires more work for the librarian may be a non-starter so you might want to thing about if there is a way to scale this sort of thing. A lot of ILSes are networked, could this tagbase be shared out and thus giving people a way to improve the system at the same time as they improve their own records? Might be a better incentive. Otherwise I sort of look at it as "Ugh our patrons are bad searchers and now their bad tags are part of my database, ughhhhh"
- Realistically I'm not sure if you could get buy in as far as having ILSes allow tags to be fed back into the system. Ideally they'd make their way into MARC but otherwise it might be worth thinking about how to add tags to ALL vwersions of an item so you'd need to deal with the work/edition issue (i.e. if someone finds the paperback of She's Not There, the tags should go to ALL copies and not just the one item UNLESS the user was searching for a format (DVD, paperback, large print) That's not at all impossible to do with technology but worth considering
- What's the system for making sure a user's typos don't become the main tags? If you have an ILS that does spellchecking/correcting in a "did you mean?" sense then people who always spell words wrong the same way may find items without ever spelling the world right. You don't want Zadie Smith's entry to have tags for Sadie Smith on them right?
- I feel like there may be an easier way to do this with data that is already out there including keywords/categories on Wikipedia and I'm not sure what else. As much as I do love the machine learning aspect of this, I wonder if your approach is the best way

In any case, I agree that ILSes suck and could be better and I applaud you working to try to make that a reality. Nice job.
posted by jessamyn at 6:12 PM on March 13, 2017