Content search not working for druid:vq627fg9932 and 18 additional druids

[caaster]

On 11/4/21, 2:55 PM, "purl-feedback on behalf of feedback@purl.stanford.edu" <purl-feedback-bounces@lists.stanford.edu on behalf of feedback@purl.stanford.edu> wrote:

Name: Emma Frothingham
Email: emma.frothingham@stanford.edu
Comment re: druid:vq627fg9932
Noticing this and some other transcripts are having some OCR issues. When you do a text search (I've been
using "interview" since I know its in the front matter) it will take you to the correct page, but the text
that's highlighted is incorrect. 
Also noticing this on the following DRUIDs: qk681gt0202, cf701vv6058, hd885gx4798, fp554yc9826, ky320sk0660,
rm620vw2642, cy927bt6518, cb635cj8418, zy935dw5016, tv486ws8292, st067nr2181, yq520nc5196, vc245bc2056,  
fr376sh7526, cc039qv8268, wg274vq1218, mt661sw7493, wd772fk6025

---

In a discussion with @anarchivist, they indicate this seems to either be a bug in ABBYY, or a bug in content search. It is noted that for druid:vq627fg9932, the ABBYY-generated page size in the XML differs from the actual page size found in the technicalMetadata, hence the incorrect display of the hit-highlighting.

@calavano do you have any thoughts here? Mark says this is not about Alto 3.1.

[anarchivist]

To clarify, for https://purl.stanford.edu/vq627fg9932 for example the image size as reported for the first page:

• in the public XML and by the IIIF server is 1700x2200
• in the ALTO file as 3400x4400

I haven't looked at all the druids provided by Emma, but I'm seeing the same thing on qk681gt0202.

[andrewjbtw]

These items were accessioned around 2019 and while I haven't checked all of them, it does not look like the files have ever been changed after accessioning. So they've had mismatching ALTO and image dimensions for a couple of years. Since the SDR accessioning process doesn't change the image size, I wonder if something happened during the OCR process where a set of images were OCR'd and then the images re-sized post-OCR and pre-SDR?

Looking at more recent oral history accessions, I don't see the same image dimension mismatches so it doesn't look like a current, ongoing problem.