[annotator-dev] Annotator and yuma-js media annotation toolkit?

Thu Jan 19 08:43:39 UTC 2012

Hi!

< Is there any reason why we couldn't use cancellation of mouse events to drag a custom selection overlay over the page? Then we could look at elementFromPoint() of the top-left and bottom-right points in this rectangle and do a search for the least common ancestor DOM element. >

Have to admit I overlooked elementFromPoint() until now... But yes, that's sounds like a very interesting idea! (In fact Yuma used to have a click-and-drag rectangle selection in an early stage. I scrapped it in favor of the current "Flickr-style" selection tool since it seemed to be more intuitive (and/or familiar) for images. For annotating parts of the entire Webpage, however, this might be much more suitable!)

Regarding the store, we're still rather conventional ;-) It's JSON(P) + some Java-based infrastructure on the server, with a relational database in the back. (I'm running a Postgres on our demo server.) The annotation model is translated into an object model. The obvious disadvantage of this is that we'll need to modify the code whenever we change the model, and there's always some tweaking needed for (de)serialization of the model classes. On the other hand, though, it's quite easy for us this way to generate alternative serializations on the fly. There's some ongoing work on an RDF serialization (see http://dme.ait.ac.at/yuma4j-server/timeline), and the old prototype used to have RSS output as well (which I can hopefully move over to yuma.js in the following weeks). 

<but CouchDB can render templates to generate RDFa for search engines and dumb clients.>

Interesting! In general, I'm really interested in NoSQL storage & and the whole CouchApp concept sounds extremely exciting!

Cheers,
Rainer

-----Ursprüngliche Nachricht-----
Von: Randall Leeds [mailto:tilgovi at hypothes.is] 
Gesendet: Mittwoch, 18. Jänner 2012 22:39
An: Simon Rainer
Cc: Andrew - FinalsClub; annotator-dev at lists.okfn.org
Betreff: Re: [annotator-dev] Annotator and yuma-js media annotation toolkit?

On Wed, Jan 18, 2012 at 11:00 AM, Simon Rainer <Rainer.Simon at ait.ac.at> wrote:
> Hi Andrew,
>
> that sounds great! I'll stay tuned.
>
> <I wonder what more specifics you have in mind for interoperability.>
>
> Still need to dive into the details of Annotator to be able to tell 
> that, I guess. But the main thing that occured to me is that Yuma 
> (=working title for our tool) might be able to fill a niche where only 
> parts of content embedded in a Web page should be annotated; in 
> particular when it's about more "exotic" media types. (We already have 
> support for maps, zoomable images, and audio/video is in preparation.)

I looked at the Yuma demos and it's great. One thing about the current Annotator storage is that it extracts the text under the selected area and uses native selection I think, whereas Yuma uses the mouse events to drag a box over the page. Is there any reason why we couldn't use cancellation of mouse events to drag a custom selection overlay over the page? Then we could look at elementFromPoint() of the top-left and bottom-right points in this rectangle and do a search for the least common ancestor DOM element. Anyway, alternative selection behaviours and selection of rich media is a topic that interests me.

>
> Also, I thought there might be a chance to join forces somehow regarding server-side annotation storage. Our goal was always to make it easy for people to set up their own storage server, integrated with their own environment, login infrastructure, etc. At the same time, we've also been having the vision of a central hosting service for people who can't or don't want to set up their own solution (or who'd rather participate in a larger "social annotation network"). While I feel there are a few differences in how we approach things, we seem to share the fundamental ideas.
>

I'm working on a new store right now. The annotator store plugin today is JSON and the reference implementation is CouchDB + Werkzeug.
However, the data model I'm aiming to work with is richer than the one used by the store plugin today. I'm tempted to go down the JSON-LD route, which seems to be gaining momentum, and exploring the current work on JSON-LD <-> RDFa conversions that have good fidelity. I also want to wrap up an out-of-the-box solution that provides a basic annotation server. The vision (tentative) is that clients that wish to can consume the JSON directly, but CouchDB can render templates to generate RDFa for search engines and dumb clients. Even if it doesn't have advanced directory capabilities built-in it might IMO provide a great solution for personally hosting annotations if I can make this a pure CouchApp.

> But there are certainly other aspects I haven't thought about yet. 
> Maybe we can even identify some low-hanging fruit for short-term 
> cooperation, who knows. But in any case: it's always good to 
> coordinate and join forces, so I'm happy get into discussion! :-)

That's the first I've chimed in on this list. I hope I'm not hijacking the thread. Looking forward to playing with all this stuff.

-Randall