BHL
Archive
This is a read-only archive of the BHL Staff Wiki as it appeared on Sept 21, 2018. This archive is searchable using the search box on the left, but the search may be limited in the results it can provide.

QA Procedures

Back to QA page

QA (Quality Assurance) Procedures


QA sampling chart

Process for spot checking scan quality of returned items.

After receipt of book trucks, original invoice is matched against shipping manifest to ensure all items sent are on the return cart. It items are missing, scanning center is contacted and provided with list of missing (or excess) items.
Document missing/excess items. Sometimes a missing item will come back on a later shipment, but sometimes it won't. Invoices/manifests are stored, reconciled at next shipment.

QA is done on a statistical sampling of books in the returned shipment, using either pdf or flip book on the IA site, based on the number received. We have a chart detailing how many books in a given shipment should be QA'd. This sampling procedure is based on a NISO standard, and is also used by IA. In general, a shipment will fail QA if more than 2% error rate is found. There is no documented method for choosing which books are part of the sample. We often do every 4th book or so, making sure to check at least one or two that look like they may be problematic (foreign language, bound-with monograph, etc.).
Though not necessary, when a cart fails QA in a spectacular way (e.g., 15% or more error rate in the statistical sample) SIL has then done further QA, first on every other book, and in some cases on the entire cart.

General procedures for checking item-in-hand against IA scanned files:

Metadata check: Make sure metadata reflects item in hand. If the metadata is attached to the wrong item, notify scanning center and they should be able to fix the problem. If QA is being done on material that was scanned more than a month ago, metadata should be corrected by hand in the portal via the editing tool.
Scan check: This is simply a matter of checking each page of the physical book while clicking along with the flip-book. Turn a page, click the scan. repeat for entire volume. It doesn't take as long as you might think, even very thick volumes can be done in about 5minutes or so. Look for:
missing pages, readable text, text that has been cut off, the scanners hand turning the page in the shot, anything that the camera might have caught that obscures the text.
DO NOT depend on pagination. Page numbers often lie or hide plates within otherwise numerically correct spans.
For any pages that seem too light, off-color, etc. download and check those pages in the pdf. Rarely, it does happen that pages missing (or seeming too light to read on the flip-book) will, in fact, be readable and well, there, in the PDF. Ensure that the OCR'd text in the PDF for suspect pages is there and more or less correct (it is OCR....)
Bottom line: if there seems to be something wrong with the flip-book double check the PDF against the book in hand page by page.

Criteria for returning a book for re-scan (or page-insertion):
Missing pages (except missing blank pages)
Text that is too light to be read by OCR (This is one that you always have to verify in the PDF. IF you can highlight the text in the PDF, the computer can read it and its dark enough. If you can't highlight, the machine can't read it--rescan)
Missing, or 'chopped off' or otherwise obscured text, figures, graphs or plates if it is part of the intellectual content of the book itself (SIL does not send back books where the ads have been cropped).