W3C home > Mailing lists > Public > semantic-web@w3.org > October 2018

Re: Semantic Web Interest Group now closed

From: Joshua Shinavier <joshsh@uber.com>
Date: Thu, 18 Oct 2018 08:50:18 -0700
Message-ID: <CAPc0OuuXZK1B-u07M9Yty-SrvMBkucOt-zmqCPwK4s4bR-ZiFw@mail.gmail.com>
To: henry.story@bblfish.net
Cc: nicolas.chauvat@logilab.fr, frans.knibbe@geodan.nl, semantic-web@w3.org
"Hyperdata" certainly has a nice ring to it, and also has some continuity
with hypergraphs in addition to hyper-text. In defense of "Semantic Web",
however, I think it is worth emphasizing that word "semantic" in certain
contexts.

To illustrate, I really enjoyed reading the trip reports from the recent
Dagstuhl seminar on knowledge graphs, including some notes by Paul Groth on
"knowledge graphs as a communication medium" that really resonated with me.
Although both a "knowledge graph" and an ML model are full of data, only in
the knowledge graph is the data entirely symbolic. This makes the contents
of a knowledge graph directly accessible -- even self-explanatory -- to
humans in a way that a big, shimmering neural network is not. If ML and
SemWeb were branches of psychology, ML would be more like behaviorism, with
its focus on inputs, outputs. and conditioning, whereas a triple store or
graph database embodies a traditional philosophy-of-mind approach. As a
close approximation of the symbols and beliefs that we assume to exist in
the minds of our fellow human beings, in a "semantic" knowledge base you
can actually look inside and see the statements.

In other words, data may very well be "machine readable" without having
"semantics", and to add to what Frans said, it's the semantics that makes
the data *human* readable.

Josh


On Thu, Oct 18, 2018 at 7:44 AM Henry Story <henry.story@bblfish.net> wrote:

>
>
> On 18 Oct 2018, at 16:23, Nicolas Chauvat <nicolas.chauvat@logilab.fr>
> wrote:
>
> Hi,
>
> On Thu, Oct 18, 2018 at 02:32:29PM +0200, Frans Knibbe wrote:
>
> To me, the term 'machine-readable web' excludes an important target group:
> us humans (assuming all current list subscribers are human). When I try
> explain the concepts of Linked Data or the Semantic Web to an uninitiated I
> like to use the terms 'web of data' or even shorter 'data web'. For some
> more finesse the term 'the web of self-explanatory data', could be
> considered, but I think just putting the terms 'web' and 'data' together
> sets the tone well enough.
> Also, a term like 'data web' is happily free of technicalities.
>
>
> +10
>
>
> I like the word hyper-data, as it helps both
> 1) show the continuity it has with hyper-text
> 2) has a fascinating relation to what is know as hyper-systems as
> described
>    Rutten, J. J. (2000). Universal coalgebra: a theory of systems.
> *   Theoretical computer science*, *249*(1), 3-80.
>
> http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.159.2020&rep=rep1&type=pdf
>
> which is a function of the form S -> 2^2^S
>
> ie one that takes you from a state to a set of sets of such states. Sets
> of states are
> often thought of as propositions, and sets of sets of those sounds like
> quad stores.
>
>
>
> perhaps even a corner stone like RDF might be replaced by something that is
> considered better in the future?
>
>
> I heard of https://ipfs.io/ lately.
>
>
> IPFS is a protocol, so in the realm of coalgebras, whereas RDF is a data
> format and
> so in the realm of algebras. Both are orthogonal, and defined in dual
> categories.
> RDF is defined in terms of IRIs and so could easily describe states on a
> IPFS System,
> just as it can on an HTTPS one.
>
>
>
> --
> Nicolas Chauvat
>
> logilab.fr - services en informatique scientifique et gestion de
> connaissances
>
>
>
Received on Thursday, 18 October 2018 16:16:17 UTC

This archive was generated by hypermail 2.4.0 : Tuesday, 5 July 2022 08:45:56 UTC