Re: Correction Re: Knowledge Graphs: the book and the hypermedia

Hi everyone,

First of all, since he was too modest in his mail, huge credit is due to 
Antoine for generating the HTML version at https://kgbook.org/! It was a 
lot of work, but his efforts have clearly paid dividends. :)

Much credit to all of the authors for their contributions and comments 
on the diverse topics! There were many interesting sub-discussions on 
specific topics while preparing the paper.

Thanks, also, to the people involved in the Dagstuhl seminar on 
Knowledge Graphs (where there were many inspiring discussions!), and 
those who provided us feedback on various versions through various channels.


@Sarven, interesting questions! I can only promise responses, not answers.

On 2021-11-19 13:42, Sarven Capadisli wrote:
> On 19/11/2021 15.58, Hugh Glaser wrote:
>> Sarven, I think you know it doesn't.
>>
>> It would be good if you just make your points in a single email, for 
>> simplicity, so our email boxes don't keep getting hit.
>> And then the discussion might even get on to an interesting topic, 
>> such as what such a graph might look like.
> 
> Should the knowledge graph of Knowledge Graphs exist?

It's something that was proposed and discussed a bit among the authors, 
but the discussion stayed "high level". It's an intriguing idea ...

> What should be able to access and interpret it?

Linked Data clients/applications, I guess. And through them, anyone on 
the Web interested in the book.

> How should the authors react to the claims made about their work?

I understand your question goes in a different direction (will get to 
that), but right now, they can send us a mail, reply to a discussion on 
the mailing list here, or add a Github issue to:

 https://github.com/Knowledge-Graphs-Book/HTML-Book/issues

> How can the accuracy of the assertions made about the work be checked or 
> compared?

Not sure I understood this question fully, as to whether the 
verification would be done automatically, or collaboratively/manually.


Regarding automatic verification, on one extreme, the most thorough 
approach would be to formalise the claims and then verify them using a 
proof assistant such as Coq. However, I don't think any claims we make 
are of the mathematically precise nature that would permit such an 
encoding. Part of CS is mathematics, but a large part is not. Also, as a 
book, our work is largely tertiary. We don't really report new technical 
claims (though we *do* make "meta-claims", in some sense; however, 
mostly not the type one could verify in Coq et al.).


In terms of collaborative, decentralised, manual/human verification, I 
personally think that in that direction lies a promising alternative to 
the classical peer review system, perhaps something like a modern 
"Webby" version of Knuth's reward cheque:

 https://en.wikipedia.org/wiki/Knuth_reward_check.

Of course, his work could only receive such a thorough 
revision/verification because of the interest it generated, but this 
also seems fitting in a way, as these are the works that tend to be most 
important to verify. Rather than sending a cheque, the book could be 
"annotated" with a note crediting the person (if they so wish) for the 
correction. Even if no change is made by the authors, it would be useful 
for readers to view comments or questions that others have as an 
"overlay", right there, in the content it pertains to.

(I know what I'm describing to you is not new, but rather largely 
related to dokeli, as you have been working on.)

Though there are of course issues to iron out (technical ones, 
scientific ones, social ones, etc.), this, to me, is an intriguing 
alternative to the classical publication / peer-review system that we 
have right now: a system that suffices, in some sense, but a system 
whose limitations slowly but surely reveal themselves over time.


Overall, in terms of what the knowledge graph of "Knowledge Graphs" 
could be useful for, I think it would be great if somehow the book could 
become "decentralized", with people leaving corrections, comments, or 
contributing related materials, in their own "webspace", as their own 
contributions, under their own authorship, pointing to the book, and 
maybe having the book point back at their content.


For example, maybe someone uses a graph from the book for a talk, or 
creates a video on YouTube, or creates some related exercises, or makes 
a Jupyter notebook to interactively go through some examples, or indeed 
maybe they provide criticism on certain parts, or want to mention some 
important works that they feel are missing. Somehow being able to both 
point to, and draw in, such content, while maintaining proper 
attribution, and without necessarily requiring an "okay" from the 
authors, is an intriguing idea.


Though Antoine has done a tremendous job with https://kgbook.org/, we 
know it can be pushed further, yes. And the ambition is there. But at 
this point in time, with the finite time/energy we could invest, this is 
what we have to offer. :)

How, specifically, it could be pushed forward in future, is an 
interesting question/discussion.

Best,
Aidan

Received on Saturday, 20 November 2021 00:44:27 UTC