Search Within Object Proposal

[Robert Sanderson]

In order to get the ball rolling before next week, Jon Stroop and I have prepared a strawman proposal for search within an object.

You can read it here:  https://docs.google.com/document/d/1F9-zelMCJqVjh-UMk8VeaPULBY2yRbI17X5EsrtcmSY/edit

Comments would be very welcome, or discussion on this week's call.

Rob

--

Rob Sanderson

Technology Collaboration Facilitator

Digital Library Systems and Services

Stanford, CA 94305

[Michael Appleby]

I'd suggest optional support for search within a region of the canvas.  Use cases for this would be large canvases (scrolls come to mind), heavily annotated canvases, and canvases with areas of unrelated content (e.g., a newspaper page).

-Mike

--
-- You received this message because you are subscribed to the IIIF-Discuss Google group. To post to this group, send email to iiif-d...@googlegroups.com. To unsubscribe from this group, send email to iiif-discuss...@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/iiif-discuss?hl=en
---
You received this message because you are subscribed to the Google Groups "IIIF Discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iiif-discuss...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

-- 
Michael Appleby
Software Engineering Manager
Yale Digital Collections Center
e: michael...@yale.edu

[Glen Robson]

Hi,

A couple of thoughts:

How do you specify multiple query terms? would it support “quoted text”? Case sensitivity?

Also we have a use case for Newspapers where we would like to return both the coordinates for the article and the coordinates of the words found. For example:

http://welshnewspapers.llgc.org.uk/en/page/view/3802098/ART86/jones%20evans

It also shows multiple word matches on the same page which would be ideally grouped in some way so the viewer wouldn’t have to go through the full result set to find all the matches on a page it is displaying.

Thanks

Glen

[Benjamin L Albritton]

To add to the good thoughts from Glen and Michael:

* How might searches on non-Latin characters be handled - and, from the manuscript realm, particularly Greek, accented, or Old English (or other similar) characters?

Thanks,

Ben

---

From: "Glen Robson" <glen....@gmail.com>
To: iiif-d...@googlegroups.com
Sent: Tuesday, October 14, 2014 4:18:20 PM
Subject: Re: [IIIF-Discuss] Search Within Object Proposal

[Stern, Randy]

Rob,

 

A question I have has to do with what sorts of text you anticipate being embedded in a canvas? For example, Harvard has a full text index of many of the page turned objects that have OCR text associated with page images. We would not necessarily expect to replicate the full page OCR text into the canvas – that seems like it would just bloat and slow down canvas creation and display. Or is that what you recommend as best practice for sharing the full document out through the presentation API?

 

Randy

[Michael Appleby]

* How might searches for/in content other than plain text be accommodated?

Michael

-- 

Michael Appleby
Software Engineering Manager
Yale Digital Collections Center
Yale University

e michael...@yale.edu
p 203.737.7255

[Tom Crane]

In the response, could the chars property convey the context of the result in a richer way?

For example in the Internet Archive book reader, a search will show results in context, on the bar at the bottom:

http://www.archive.org/stream/birdbookillustra00reedrich#page/238/mode/2up/search/hawk

This is driven like this in the JSON:

"text": "...The nest is of the usual {{{Hawk}}} construction and the two to four..."

full results:

https://ia902303.us.archive.org/fulltext/inside.php?item_id=birdbookillustra00reedrich&doc=birdbookillustra00reedrich&path=/17/items/birdbookillustra00reedrich&q=hawk&callback=br.BRSearchCallback

If the server supports richer search features, lemmatization and so on, it offloads the work of working out what to highlight, which might not be as simple as string matching the q request parameter in the chars response. I'm not sure I like the {{{...}}} though.

For the Wellcome Library we approached this slightly differently, with a "before" and "after" to give context (the size of these values also controllable through parameters):

http://wellcomelibrary.org/service/search/b18035978/0?t=tangled

[Simeon Warner]

I like the "before" "word" ("match"?) "after" better than {{{ }}}. This
seems easier to handle and has no escaping issues (what if {{{ really
occurs in the text...?).

Cheers,
Simeon


On 10/15/14, 12:25 PM, Tom Crane wrote:
>
> In the response, could the *chars *property convey the context of the

> --
> -- You received this message because you are subscribed to the
> IIIF-Discuss Google group. To post to this group, send email to
> iiif-d...@googlegroups.com. To unsubscribe from this group, send
> email to iiif-discuss...@googlegroups.com. For more options,
> visit this group at https://groups.google.com/d/forum/iiif-discuss?hl=en
> ---
> You received this message because you are subscribed to the Google
> Groups "IIIF Discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to iiif-discuss...@googlegroups.com

> <mailto:iiif-discuss...@googlegroups.com>.

[Tom Crane]

Something else to think about - here the highlighted rectangles don't overlap, but the "context" overlaps:

http://wellcomelibrary.org/service/search/b18035978/0?t=dispassionate

In action: http://wellcomelibrary.org/player/b18035978#?asi=0&ai=266&q=dispassionate

We're not making use of the context in the UI yet - but does this overlap even matter? Maybe leave it up to the client to decide whether to do anything about it. It wouldn't really matter in the Internet Archive's result UI.

Thinking about it maybe this is a red herring, and it doesn't matter at all.