Finalize a strategy to start and finish the project i.e. a timeline.

[Amanpreet Singh]

Current proposal can be found on https://www.mediawiki.org/wiki/User:Apsdehal

[Simone Fonda]

As briefly discussed yesterday, here are a few suggestions for the proposal:

- we might want to add some more detailed and practical use case which
we aim to solve (eg: add statements to mediawiki, tracking the user
and source)
- we might want to add some more formal requirements the project
should fulfil (eg: help the user in avoiding making the same statement
twice; Pundit should authenticate mediawiki users; users should be
able to annotate any web site; pundit should get configured with
mediawiki predicates and items; pundit should be packaged into a
bookmarklet and/or a browser plugin/extension)
- we might want to add some words about the expected data flow.
One-directionally: user creates an annotations, it gets saved in the
annotation server, it gets "transported" on mediawiki
- we might want to declare precisely unclear sides of the project
which needs to be analysed in depth (eg: how to move annotations from
pundit to mediawiki; how to authenticate mediawiki users; how to feed
pundit with mediawiki predicates/items; privacy policies of mediawiki
regarding users)
- we might want to add future steps which this project would allow to
get developed in some other future project (eg: richer annotations,
bidirectionality of data flow between pundit and mediawiki, etc)


Im pretty sure i'm forgetting something so please, add your comments! :)

Simone

[Amanpreet Singh]

On Thu, Feb 20, 2014 at 2:57 PM, Simone Fonda <fo...@netseven.it> wrote:

On Wed, Feb 19, 2014 at 2:23 PM, Amanpreet Singh
<amanpreet...@gmail.com> wrote:

> Current proposal can be found on
> https://www.mediawiki.org/wiki/User:Apsdehal

As briefly discussed yesterday, here are a few suggestions for the proposal:

- we might want to add some more detailed and practical use case which
we aim to solve (eg: add statements to mediawiki, tracking the user
and source)

These all statements have to be clarified how are we going to track the user, I think best would be to 

use WIkidata api or an oauth by mediaiwiki. Currently there are two use case listed in the proposal.

 

- we might want to add some more formal requirements the project
should fulfil (eg: help the user in avoiding making the same statement
twice; Pundit should authenticate mediawiki users; users should be
able to annotate any web site; pundit should get configured with
mediawiki predicates and items; pundit should be packaged into a
bookmarklet and/or a browser plugin/extension)

I think making ajax call to test what user is typing would be great to reduce repetition. I think it would be based to

package the whole project as a mediawiki extension 

- we might want to add some words about the expected data flow.
One-directionally: user creates an annotations, it gets saved in the
annotation server, it gets "transported" on mediawiki

This also needs to be discussed as we are not clear where we are going to store annotations till now.

Needs a group chat for consideration. 

- we might want to declare precisely unclear sides of the project
which needs to be analysed in depth (eg: how to move annotations from
pundit to mediawiki; how to authenticate mediawiki users; how to feed
pundit with mediawiki predicates/items; privacy policies of mediawiki
regarding users)

Yes this job is to be done, I will add these to proposal asap. 

- we might want to add future steps which this project would allow to
get developed in some other future project (eg: richer annotations,
bidirectionality of data flow between pundit and mediawiki, etc)

This is also a good option to diversify the project scope. 

Im pretty sure i'm forgetting something so please, add your comments! :)

We still have to discuss the GUI part, thats' important :D 

Simone

--

Amanpreet Singh,

IIT Roorkee

[David Cuenca]

If, as Simone says, it should be possible to annotate all pages then it is crucial to start with a bookmarklet/browser plugin/extension. As an example I recommend taking a look to [1] since they are following the same approach.

A possible workflow could be as follows:

1) with the extension/bookmarklet installed the user selects the text to annotate

2) entity extraction to recognize possible subject-object

3) in the Pundit servers it is saved all the information, including text location, selection, etc, An ID for the annotation is generated (i.e bb5eb045)

4) If by any chance the annotation is a valid Wikidata triple (and not a comment, or a connection between two text fragments), then the information is transfered to the wikidata item (triple, plus source: origin URL, date, etc see [2]), additionally a new property should be created to refer to the annotation saved in the Pundit server (for instance, it could be something like "annotation URL:www.thepund.it/id/bb5eb045"). By clicking this annotation ID it should be possible to show it in context (text highlighted, with the Pundit activated)

5) if the page being annotated is a Wikisource page, then the item representing that book should be used as source information in wikidata.

AFAIK, 2 and 3 are already done by Pundit, about 1 no idea, and 4 and 5 should be the main aim of the project.

A possible extension to the project could be to use a Mediawiki installation to save all annotations instead of relying on the external Pundit server (cfr. [3]). And of course, bidirectionality would be a nice add-on :)

Cheers,

Micru

[1] https://hypothes.is/alpha/
[2] https://www.wikidata.org/wiki/Help:Sources#Web_page

[3] https://www.mediawiki.org/wiki/Extension:Annotator

--
You received this message because you are subscribed to the Google Groups "Annotation tool GSoC" group.
To unsubscribe from this group and stop receiving emails from it, send an email to annotation-tool-...@googlegroups.com.
Visit this group at http://groups.google.com/group/annotation-tool-gsoc.
For more options, visit https://groups.google.com/groups/opt_out.

--
Etiamsi omnes, ego non

[Simone Fonda]

Hey David!

Nice to hear from you again!! :)

On Sat, Feb 22, 2014 at 1:10 PM, David Cuenca <dac...@gmail.com> wrote:

> 1) with the extension/bookmarklet installed the user selects the text to
> annotate

Creating a bookmarklet is not a problem. A browser extension is
somewhat more delicate, but we already have some very solid proof of
concept for firefox and chrome, if not beta versions.

> 4) If by any chance the annotation is a valid Wikidata triple (and not a
> comment, or a connection between two text fragments), then the information
> is transfered to the wikidata item (triple, plus source: origin URL, date,
> etc see [2]), additionally a new property should be created to refer to the
> annotation saved in the Pundit server (for instance, it could be something
> like "annotation URL:www.thepund.it/id/bb5eb045"). By clicking this
> annotation ID it should be possible to show it in context (text highlighted,
> with the Pundit activated)

How do you envision this to happen? The annotation client talks
directly with the Wiki* servers, or wiki* servers query the pundit
server for annotations?

About the second part (clicking the annotation id brings you to the
context), it should be possible with a browser extension, not with a
bookmarklet, since pundit doesnt have any control on the web page.

Best,
Simone

[David Cuenca]

On Mon, Feb 24, 2014 at 2:20 PM, Simone Fonda <fo...@netseven.it> wrote:

Hey David!

Nice to hear from you again!! :)

A bit lethargic lately, but still alive :)

 

Creating a bookmarklet is not a problem. A browser extension is
somewhat more delicate, but we already have some very solid proof of
concept for firefox and chrome, if not beta versions.

Then for the GsoC I would leave it only as a bookmarklet, just as a proof of concept. The extensions can be a later project, or if there is some extra time left.

 

How do you envision this to happen? The annotation client talks
directly with the Wiki* servers, or wiki* servers query the pundit
server for annotations?

I think it should be either the annotation client talking with the Wiki servers or even the annotation server through oauth.

 

About the second part (clicking the annotation id brings you to the
context), it should be possible with a browser extension, not with a
bookmarklet, since pundit doesnt have any control on the web page.

I was thinking something like the toolbar in reddit.com (Preferences->display links with a reddit toolbar), where a link is generated that loads a toolbar plus the original page, and for that you don't need to have any extension installed. It is open source, so maybe you could take a look to their github to see how it's done?

Cheers,
David

[Cristian Consonni]

2014-02-24 15:17 GMT+01:00 David Cuenca <dac...@gmail.com>:
> A bit lethargic lately, but still alive :)

(Well, it is good to hear that lethargy is widespread ;-) )

>> Creating a bookmarklet is not a problem. A browser extension is
>> somewhat more delicate, but we already have some very solid proof of
>> concept for firefox and chrome, if not beta versions.
>
>
> Then for the GsoC I would leave it only as a bookmarklet, just as a proof of
> concept. The extensions can be a later project, or if there is some extra
> time left.

+1

>> About the second part (clicking the annotation id brings you to the
>> context), it should be possible with a browser extension, not with a
>> bookmarklet, since pundit doesnt have any control on the web page.
>
>
> I was thinking something like the toolbar in reddit.com
> (Preferences->display links with a reddit toolbar), where a link is
> generated that loads a toolbar plus the original page, and for that you
> don't need to have any extension installed. It is open source, so maybe you
> could take a look to their github to see how it's done?

+1

2014-02-22 12:24 GMT+01:00 Amanpreet Singh <amanpreet...@gmail.com>:
>> - we might want to add some more detailed and practical use case which
>> we aim to solve (eg: add statements to mediawiki, tracking the user
>> and source)
>
> These all statements have to be clarified how are we going to track the
> user, I think best would be to
> use WIkidata api or an oauth by mediaiwiki. Currently there are two use case
> listed in the proposal.

i have rewritten the use cases because thy were to generic, IMHO.

Cristian