Use with content search 2.0

[fitnycdigitalinitiatives]

Hello,

I'm wondering if you have any example code to take solr responses and turn them into in Content Search 2.0 compliant responses. The code you have here was very helpful to set up a working 1.0 response, but I'm having trouble figuring out 2.0. It's a bit more complicated, specially how the 'hits' of 1.0 correspond to 'annotations' in the response.

Thanks,

Joe

[jbaiter]

Hey Joe, I have not yet looked into implementing Content Search 2.0 on top of the plugin, but from looking at the spec, especially "4.3 Extended Responses" (and especially the sample in 4.3.1), it seems that:

• items contains Annotations with the exact location of the matching terms/phrases. This would be what's inside of the highlights data structure in the plugin response
• annotations contains an AnnotationPage with a list of Annotation resources that have a motivation: "contextualizing" key. On these annotations, you'd link to your match annotation in items via the .source key, and then provide prefix and suffix information about the context of the match. You can get this with a bit of string wrangling from the text field in the plugin snippet response, by looking at the <em>...</em> spans, matching them up to your highlights and then using that to generate prefix/suffix information. Much like was done in the 1.0 implementation: https://github.com/dbmdz/solr-ocrhighlighting/blob/main/example/iiif-prezi/main.py#L85-L93

[fitnycdigitalinitiatives]

Ok, that's helpful. One more question, in your code a single 'hit' could reference back to multiple annotations/resources because of anno_ids = [], but how would that then work in 2.0? Would those be separate items? Because the source in 2.0 isn't a list like hits.annotations could be a list? I hope that makes sense.

Also, do you know about the id of the 'item'? Does that refer to something specific or can it just be something sequential, ie match-1, match-2, match-3 etc.?

[jbaiter]

Ok, that's helpful. One more question, in your code a single 'hit' could reference back to multiple annotations/resources because of anno_ids = [], but how would that then work in 2.0? Would those be separate items? Because the source in 2.0 isn't a list like hits.annotations could be a list? I hope that makes sense.

Excellent question, I can't really tell from the spec how that's supposed to be handled. This would be in cases where you search e.g. for a phrase and get a match that spans multiple distinct rectangles on the page (e.g. due to a column or line break). Since IIIF Annotations are based on W3C Web Annotations, the way to go would probably to use an array for selector that has every rectangle associated with the match. But I don't know how well viewers support that.

Also, do you know about the id of the 'item'? Does that refer to something specific or can it just be something sequential, ie match-1, match-2, match-3 etc.?

I think these can be arbitrary, you could also use UUIDs or something like that. They don't have to be dereferencable, especially in cases like ours where we generate the annotations on-the-fly.

[fitnycdigitalinitiatives]

Ok, thanks so much for your help. I appreciate it.

And thank for the plugin which has been working great for us.