kwarc / kat Goto Github PK

It would be nice if there was a help button in the annotation bar. That might lead to a FAQ or something (which would have to be written; maybe on the github wiki).

The function KAT.storage.gui.getRange(...) doesn't seem to include the math expressions.

You can't reach the menu at the very bottom since it goes out of the screen.

We could possibly use something like doT

I have followed the installation instructions in readme.md, but when I annotate a symbol and enter a symbol name, clicking on save does not work, I can only cancel my way ouf of the dialogue. (this may be a consequence of using --force in starting the grunt server.

We are currently not making progress on CorTeX, therefore we should think of other venues for KAT integration. A natural one is integration into MathHub, where we can use KAT with the OMDoc KannSpec for semantic annotation.
This is the KAT-side tracker bug, to be used for planning. The MathHub-side one is KWARC/MathHub#164 cc: @m-iancu @LiftnLearn

Here are the things that need doing (from the KAT side things are relatively simple)

• document how to integrate the KAT js into an arbitrary web server.
• make the OMDoc KannSpec more complete.
• specify and document the RDF storage procedure.

I guess we should GPL-license KAT. There are lots of good templates around.

After successfully making the first annotation in a page, the next click on "Annotate" simply scrolls the page to the top.

The issues only appears some of the time, mostly on larger selections (e.g. an entire paragraph). And I still had to dance around issue #8 so they might or might not be related.

I am using FF 27 on Ubuntu 13.

possibly we just have to re-tokenise the sample document

The KannSpec stanford.xml was never complete in the RDF templates, and so the new version (see #15) does not validate, since we do not have rdfpred attributes on the <field> elements. These need to be systematically added (the ontology at http://acoli.cs.uni-frankfurt.de/resources/olia/stanford.owl should have the respective relations).

It would be good to have a "report issue" button in the annotator bar. It should just link to the issues page on github. Then our users (with Randa we may have one) can more easily report issues.

Clicking on it should display general information about the field (i. e. the contents of the tag) and optionally validations it has to pass.

I'm not sure what it should do, but at the moment it seems to rerender the whole page.

The current workflow of: "open browser, download text, insert KAT, annotate" is tedious and inefficient if you would have a corpus with multiple similar articles.
I was thinking of integrating KAT into a Firefox extension (I've worked a lot with Chrome extension, but I know FF is very similar so it shouldn't be a problem) which would be a permanent button in your browser. Therefore whenever the user would want to annotate a document they would click the button and the system would be inserted on top of the current webpage.

When you want to open/close the sidebar by hovering over the bar and have an edit/create-field open, then the wrapper seems to block the sidebar where the field is on a vertical level.

We want to read RDF XML back in, that means we cannot use the templating engine to generate the XML. Instead we decorate

• the <concept> element in a KannSpec with a rdftype="o:foo" attribute, this generates a RDF triple of the form <rdf:type rdf:resource="o:foo"/>.
• the <field> element with a ```rdfpred="o:bar" attribute, this generates an RDF triple
  • <o:bar>baz</o:bar> if the type of the field is text
  • <o:bar rdf:resource="#ref#"/>, if the type of the field is reference and #ref# the value of the reference.
  • <o:bar>#val#</o:bar> if the type of the field is select and #val# is the value of the chosen option.
  • <o:bar rdf:resource="#obj#"/> if the type of the field is select and the chosen option carries an attribute rdfobj="#obj#".
    This should cover all the cases.

Currently, there is a serialization method in Annotation.js that outputs JSON. For the CorTeX integration we need:

• Creating RDF instead of JSON
• Having a top-level serialization methods for convenience. E.g. I just call:

var rdf_string = service._annotationRegistry.getAnnotations(serialize:"RDF");

and that would already provide me with a serialized RDF document that I can insert back into CorTeX.

The title is "Annotator Testing", I think it should be "KAT Demo".

Getting an error:

TypeError: result[key] is undefined

sometimes when making selections. In that case not bubble to annotate shows up.

From what I am seeing at the moment, the KAT annotations are relative to the document. That works while you have a single document, but not for a corpus, where you want each annotation to point to an absolute URI for the document.

CorTeX passes that information along with the document body, so KAT should make use of it and we should decide on how. The simplest extension would be to add a setURL method to the KAT service, or something on those lines.

In the CorTeX workflow, we have many annotators working on the same corpus of documents. Hence, they also will annotate with the same ontology.

To achieve this, we need a way for KAT to pick ontologies server-side and let the client use them directly. Since I am writing the KAT-enabled page in the first place, I can add any parameters necessary to make that happen, as long as KAT is able to fetch the ontology from the server.

I would appreciate help with understanding exactly how to do that.

At the moment the annotations are shown in the order they were created, but it might be useful to show them in the logical order in which they appear in the document.

I have recently updated the code to just use the color of the last annotation created, but this should be solved in a different way. Maybe we should mix colours.

The code where the color is determined is here:
https://github.com/KWARC/KAT/blob/dev/src/KAT/storage/annotation.js#L469-L474

I executed the commands in the readme.md, but I had trouble finding out what npm is and how to install it. Say something about this or link to a good description.

It might be much easier to have a point and click feature which allows users to select the current tag by clicking on it. Generally relevant content is contained within the tag - in the case of MathML in a math tag. This can offer a much smoother and more precise workflow which would not be subject as much to the change of content.

• Implement GUI (@LiftnLearn)
  • Text Fields
  • References
  • Selects
• Implement Storing / Loading / Validation (@tkw1536)
  • Text Fields
  • References
  • Selects

I think we should think of annotation frames instead of concepts, So we should have . We should also update the manual.

We should use the element of the KAnnSpec for this. We also might have to change the templating.

We need some storage solution for KAT. The most natural way to do that is to integrate KAT into CorTeX. This also needs to be documented in the manual.

It seems like sometimes the tooltip doesn't disappear after you stop hovering over the annotation; although I'm not sure yet when exactly it appears nor if it's a KAT or JOBAD issue. Also I think there shouldn't be any tooltips shown under the sidebar when it's open.

When we have the "show all annotations" (the little eye) in the KAT control panel, then the visualization boxes of both source and target of the annotation frames should be shown .

The popover stays fixed when you scroll and there are still scenarios where you can close the sidebar and the popover stays.

Still to be done:

• When the sidebar is closed the popover is deleted but if you reopen it you have to click the button twice for the help to reappear

and reference the manual there. This now has more up to date installation instructions, so I think the ones that are there can go.

I would like to have a nice demo setup like JOBAD has. Once that is done, we should also change the demo URL in kwarcpubs.bib.

The browser window should automagically focus the annotation we're looking at.

Nathan Carter told me at CICM that he may have a library we could use for visualizing nested/overlapping annotations in KAT. They have (just) reworked Lurch (see http://lurchmath.org/) for the web, and have developed a CSS library for the bubbly visualization of their "annotations" (to be relesased in the next days apparently). @tkw1536 @LiftnLearn, this may be a good thing to look at, I will get a demo in the next days. I will keep you posted.

In readme.md I see the line

# Whenever you update you will have to re-run this command.

and it is unclear to me which command it pertains to, and when to re-run, when I update grunt or KAT?

The installation instructions in readme.md end in

take a look at the demo.

Give a few sentences that explain what to expect and what can be done, link to the paper. In particular mention

• you get a document with the OMDoc spec (link to explanation) preloaded
• annotation mode vs reading mode,
• select a phrase annotate it by clicking it again: select from type list, give more data, save
• ...

I get

Michaels-MBP:KAT kohlhase$ grunt serve
Loading "jsdoc-ng-task.js" tasks...ERROR
>> Error: Cannot find module '../node_modules/jsdoc/node_modules/requizzle'

Running "jshint:all" (jshint) task
>> 15 files lint free.
Warning: Task "jsdoc-ng" not found. Use --force to continue.

Aborted due to warnings.

but grunt serve --force works after some complaing.