orangewolf
commented
1 year ago Closed orangewolf closed 1 year ago
orangewolf
commented
1 year ago orangewolf
commented
1 year ago the catalog search and uv search are performing two different functions.
the catalog search will search for ocr and metadata. the parent uv search however searches only for the searchable_text field and the ocr. (reminder that the the child work only does the ocr search) this was my understanding on how it should work and what we discussed during yesterday's client meeting.
"register of the officers" is the title of the work. the title is a metadata field, which is why the catalog search returns it. however, none of the child works for this parent have anything in the searchable_text field (per the csv), so only ocr results should be returned here.
however, I do see where even an ocr term ("revised) is not being returned for the poe parent work, although it is being returned for the poe child work. this is not the case for all parents however because I can still search for the term "stare" on the cat parent and it returns the value.
I do also see that the left and right arrows at the top of the viewer are now on different rows for some reason. :thinking:
| poe parent | poe child | cat parent |
|---|
orangewolf
commented
1 year ago Yeah, I guess what we're trying to convey is that it seems like these materials were not OCRed.
orangewolf
commented
1 year ago I gotcha.
orangewolf
commented
1 year ago dev convo: https://assaydepot.slack.com/archives/G0313NK5NMA/p1653070746852929
orangewolf
commented
1 year ago Alisha,
I can't get into the slack link (although I do have a slack login and have access to the Samvera Community, as does Randy - for some reason this link wants us to sign in through a scientist.com Google account or get a guest account).
orangewolf
commented
1 year ago sorry rachel, that thread was for the other devs on this project. :smile: it's not in the "samvera community" workspace.
please take a look at one of the poe child works and scroll to the "items" at the bottom. in the "actions" section there are download options (which will be removed for non superadmins in #113). the "download as ocr" will show that the ocr "translation" did in fact happen. unfortunately, due to the quality of an image, lack of contrast, complexity of typeface, etc., not all images will have very accurately ocr'd text. this is a limit of the present state of that feature within the newspaper_gem.
revising that gem at this time to account for the limitations above would be out of scope of our current contract. however, any image that has text placed within the searchable_text field on a child image will return accurate results from a catalog or parent uv search. :grin:
orangewolf
commented
1 year ago Thanks. We actually still have the OCR from the original metadata export so that could be a workaround. But, I'm not seeing "download as OCR" on our dev server (and I'm not seeing any Poe child works on your staging server anymore?!?), so I'm not sure that's the issue with this item.
orangewolf
commented
1 year ago Never mind, I guess you meant "Download (as txt)" and there is text in there.
orangewolf
commented
1 year ago whoops! yes, I meant "download (as txt)".
also, below are a few links to the poe import
orangewolf
commented
1 year ago ref !116
| before | after |
|---|---|
  |
We uploaded the Poe collection. It appears it uploaded fine and sidekiq is not longer working on anything. However, strange things are occurring. The viewer's top arrows are not lined up, the main search does not work for OCR text, the viewer search doesn't work either, and we're receiving no matches errors in the viewer. I believe Rachel sent you a subset of this collection. If you need a larger subset, please let us know. Take a look...
https://user-images.githubusercontent.com/1054448/203993807-f7e5349f-6f60-4f31-aad2-073102ae06ad.mov