Hi Eric,

I really enjoyed this message. Thanks for sharing!

Best,
Lisa

> On Jun 7, 2016, at 2:49 AM, Eric Lease Morgan <emor...@nd.edu> wrote:
> 
> In the past few weeks I have had some interesting experiences with WorldCat, 
> VIAF, and the Levenshtein algorithm. [1, 2]
> 
> In short, I was given a set of authority records with the goal of associating 
> each name with a VIAF identifier. To accomplish this goal I first created a 
> rudimentary database — an easily parsed list of MARC 1xx fields. I then 
> looped through the database, and searched VIAF via the AutoSuggest interface 
> looking for one-to-one matches. If found, I updated my database with the VIAF 
> identifier. The AutoSuggest interface was fast but only able to associate 20% 
> of my names with identifiers. (Moreover, I don’t know how it works; 
> AutoSuggest is a “black box” technology.)
> 
> I then looped through the database again, but this time I queried VIAF using 
> the SRU interface. Searches often returned many hits, not just one-to-one 
> matches, but through the use of the Levenshtein algorithm I was able to 
> intelligently select items from the search results and update my database 
> accordingly. [3] Through the use of the SRU/Levenshtein combination, I was 
> able to associate another 50-55 percent of my names with identifiers.
> 
> Now that I have close to 75% of my names associated with VIAF identifiers, I 
> can update my authority list’s MARC 024 fields, in turn, I can then provide 
> enhanced services against my catalog as well as pave the way for linked data 
> implementations.
> 
> Sometimes our library automation tasks can use a bit more computer science. 
> Librarianship isn’t all about service and the humanities. Librarianship is an 
> arscient discipline. [4]
> 
> [1] VIAF Finder - http://infomotions.com/blog/2016/05/viaf-finder/
> [2] Almost perfection - http://infomotions.com/blog/2016/06/levenshtein/
> [3] Levenshtein - https://en.wikipedia.org/wiki/Levenshtein_distance
> [4] arscience - http://infomotions.com/blog/2008/07/arscience/
> 
> —
> Eric Lease Morgan

Reply via email to