(Mis)Match Detection

Is the OCLC record a true match for our original record?

• Large percentage difference in title string
• Difference of > 1 year in 008 date value
• If present, no matching ISBNs or ISSNs
• Presence of electronic resource fields in a print record, and vice versa

(Mis)Match Detection

Use pymarc and OCLC Metadata API to compare record fields

Potential mismatches

Results

976 random records, no issues

~2% error rate for minimal records

Quality Assessment

How much "better" is the OCLC record than our original?

Quality Assessment

Use pymarc to calculate Thompson-Traill scores for each pair of records

See also: https://github.com/pkiraly/metadata-qa-marc

Thompson, Kelly, and Stacie Traill. “Leveraging Python to Improve Ebook Metadata Selection, Ingest, and Management.” The Code4Lib Journal, no. 38 (2017).

http://journal.code4lib.org/articles/12828.

Thompson-Traill Scores

ISBNs

Subject Headings

Additional Authors/Contributors

Description and TOC

RDA fields

Dates

Language and country codes

Results

Local Records = 11.15

OCLC Records = 15.62