RE: new annotation software

Julie,

My comments are interspersed in your text.

-bryan

-----Original Message-----
From:	Julie Gibson [SMTP:julieg@weborganic.com]
Sent:	Wednesday, May 10, 2000 8:25 PM
To:	Bryan Thompson
Cc:	Bryan Thompson; jgarfunk@bbn.com; www-annotation@w3.org; annotate@cog-tech.com
Subject:	Re: new annotation software

Bryan,


> I see a problem with managing scale for annotation systems.  This includes the problem of managing topics, the definitions of those topics, and the specific referents in documents that serve as seeds points for annotations.  It also includes managing the aggregation and abstraction of annotations and leveraging the expressed knowledge for search and decision making support.

This makes a good checklist to compare annotation solutions against:
scale
topics and how to define them
document referents
aggregation and abstraction

yes, and also how to maintain annotation link integrity, especially for fine-grained annotations.

Our approach leaves most of these issues in the hands of the owner of the topic, in fact a topic is defined in our terms by the combination of a set of URIs with a group of interested members. The set of URIs is finite but the group can be completely public. Until I read through the Desire pages I had not thought of a method of dynamically assigning the set of URIs but I can see that could be an added on feature of our system.

Scale is going to depend on the database in terms of simple size and how many annotations are humanly useful at a seed and still leave the original document readable. We have found that we have to create regular archives for our demo page because it gets too full to be useful.

That is interesting.  This may also be a function of the annotation structure.  E.g., argument-based models vs relatively unstructured commentary models.  Most annotation systems fail to explicitly manage the organization of information, e.g., as a knowledge management activity.

Aggregation and abstraction is going to depend on the user, by storing all annotations in an XML database we assume that they will be as useful as the owner can make them With your suggestion of collecting opinions you have pointed out that the annotations could be structured to collect a limited set of feedback which could then be abstracted.

I am primarily interested in structured annotations based on cognitive models, in particular, annotation systems for supporting argument-based reasoning.  Arguments are relatively easy for people to externalize (when compared with causal models).  I see the aggregation of arguments as partly a matter of applying a consensus policy (to dynamically aggregate estimates with regard to the truth of individual propositions) and partly a matter of abstracting the view of the argument.  (Crit.org did some work on the latter with contextures in an argument-based representation.)

I just read through Libby Miller's paper on aggregating annotations using RDF, which presents yet another view on aggregation.

I'd be pleased if you would look through our planned features page and add other suggestions. So far we have found that our system will be extensible to all of these future features but we are slightly(!) limited by resources.

I would be delighted to do so.  Can you point me directly to this URL?

I am interested in creating an commercial grade open source annotation project.  It is my hope that this will make it easier for people to develop specific products and services based on annotation systems.

> As you can see, from my other post, I feel that a broader range of solutions is necessary.  I am interested, and maybe we should take the discussion of business models out of the group, in why you chose the specific business model, which is clearly tied to your assertion that

I'm not sure what you mean by our business model. Unless you use this phrase to refer to the fact that we have developed only one aspect of an annotation system. It's not that we were theorists about annotations but that we were using the web for a purpose; developed a way of carrying out that purpose (teaching to a group and using feedback from the group to add to the accessible content of the web pages) then realised that what we were doing was a form of annotations.

I see.  You have a particular problem in mind, relating to education, and developed the technology to address that problem.  I did not realize that.  I had thought that you were approaching "annotation" market more broadly and using a pay-per-seed structure as the revenue model.

This means that we keep in the front of us that original purpose and since it's been complex enough applying that limited set of requirements we have not attempted to generalise. My justification of creating seed points is a post-fact justification, from our initial application when we were collecting responses from students and hand inserting them into the HTML we realised that it was kept more readable when we collected them at points rather than randomly dispersed them throughout the page.

I think that the main role that I can perform in any attempt to create standards is to ensure that our application of annotations on the web can remain as a subset of any theoretical descriptions.

> Your other point,
>
> This makes us at an opposite pole from third voice but there is purpose
> for both approaches. At least all of our annotations can be collected in one database by the owner and made use of.
>
> is also very interesting.  I feel that there should be interchange and encryption support for annotation and collaborative systems.

will XML be able to solve this? maybe a combination of an annotation DTD with other special DTDs depending on the topic.

> I am particularly loath to create content in a database owned by a third party (e.g., ThirdVoice).  The lack of interchange standards means that I can not obtain copies of that information and limits my ability to develop intelligent behaviors that exploit the annotation base to create information or services.  Collaborative systems without open logies and interchange standards, people and businesses should be able to use public servers and have confidence in (and control over) the access control policy.

Wow you've dragged me away from our current project which is to set PageSeeder up on an Excelon system which will mean that the original documents will all be XML so that there will be a seamless storage system for both the annotations and the original document fragments (in fact this project will enable the user to exchange whole paragraphs in the original document with annotations to create a new document.)
But I'm pleased to find some movement at the station.

Cheers
Julie
-
--------------------------------------------------------------
Julie Gibson                     E-mail: julieg@weborganic.com
Weborganic Systems Pty. Ltd.     http://www.weborganic.com
Level 10, 91 York St Sydney NSW 2000 Australia
Phone +61 2 9262 4777

Received on Friday, 12 May 2000 11:57:13 UTC