[Textop-en-phil] Report about our first Skypecast

[Larry Sanger]

All,

I thought I would share my impressions with you about our discussion this
morning via the "Skypecast."  Others who were involved said (promised??)
that they would elaborate certain things they were saying in e-mails to one
or more of the lists.

In attendance were Anat Biletski, a Hobbes specialist and distinguished
professor from Tel Aviv University now visiting at Boston U.; Kunal Sen,
Executive Director, International Digital Product Development for
Encyclopedia Britannica; Philippe Martin, who researches knowledge
representation; Howard Burrows, physiologist, information theorist, and
project manager who has lived in much of our problem space; and Dave
Cormier, EducationBridges moderator who gave me useful advice about how to
manage the Skypecast.  My apologies if I left anyone out, but as you can see
it was an excellent (and very international!) group, if small.

Prof. Biletski asked about how feasible the project is, anyway.  (It was
more complicated than that, but that was the gist of it, I think.)  My
answer in brief was that there are aspects of the project that are obviously
feasible, because I've shown how to do them (e.g., chunking texts and
providing summaries of the chunks), while the feasibility of other, probably
more interesting aspects remains obviously unproven.  (I always hedged my
bets in a similar way about the prospects of Wikipedia, you know.)  I don't
think we really began to explore the question of feasibility, though.  I
think another point that was brought up in some back-and-forth involving
Philippe and Kunal and myself was the following.  Given that the Collation
Project will involve chunking small (paragraph-sized) parts of texts, rather
than encyclopedia articles or books etc., and putting them into an
*extremely* detailed outline, some of the more obvious concerns about
feasibility connected to the difficulty of dealing with vague and ambiguous
categories might be finessed.

Kunal Sen asked two excellent questions (and showed himself to possess
professional expertise on two or three completely different aspects of this
project): (1) while it is simple to map chunks onto a pre-existing
ontological (or taxonomic) structure, how do you create the ontology in the
first place?  My answer: nodes of the outline (ontology/taxonomy) are
created at least initially on a more or less ad hoc basis, just to
categorize chunks.  Nodes are *not* created "a priori," i.e., before any
chunks can be found to put into them.  Beyond that, it's (a) a practical
matter of deciding on a procedure for arriving at agreement among the
outline-builders, and (b) deciding on (or, better, elaborating and refining)
principles according to which the outline is built out.  Not that either (a)
or (b) is easy, but they are both issues I've thought a fair bit
about...and, I would add now, a main purpose of the pilot project is to
start working out solutions to these problems.

(2) For nonphilosophical or mathematical subjects, such as Russell's
Principia, how could that be chunked?  My answer was that, for a lot of
proofs in logic texts, it might end up one chunk per proof, to which Kunal
said that that is what he thought, but that then the chunks might end up
being pages long, because the proofs are sometimes pages long.  I then
suggested that sometimes it will be possible to identify subproofs, or
discussions of particular premises, which can be placed at subheadings
underneath the heading for the main argument.  (This did not, I see now,
answer his main worry, that taking proofs apart in this way would render the
original argument hard to follow.  But I don't think so: proofs that require
pages generally speaking have typical *parts* that are discussed by other
theorists, i.e., subproofs and key premises that are discussed at great
length by many different theorists who present their own versions of the
proof in question.)

Howard Burrows and Kunal Sen also discussed the similarity and differences
between the Collation Project and Mortimer J. Adler's *Syntopicon*.  Howard
urged me to read Adler's introduction to the Syntopicon, which is very
relevant to our endeavor.  I think we agreed that the main difference
between the Collation Project and the Syntopicon is that the latter is much
more coarse-grained in two ways: its outline is not as detailed as ours will
be, and it indexes sections of works, rather than individual sentences and
paragraphs.  (And, of course, the Collation Project will be digital and use
lots more books!)

Toward the end of the call, there was a (for me) somewhat
difficult-to-follow exchange by two very subtle thinkers who have (it seems)
different approaches to what we're doing, Philippe Martin and Howard
Burrows.  Philippe defended his notion that an *extremely* fine-grained
approach to (some of) Textop's projects is possible, while Howard emphasized
some conceptual difficulties inherent in interpretation and classification.

One thing that came out of that last exchange was that there is in principle
nothing wrong with making initial outline nodes "placeholders" ("placing a
flag in the sand" I think someone said) useful for an initial
categorization, not bound by any strict rules of ontology.  My own
conclusion was that, if the outline ever can be made to follow some formal
or semi-formal rules of ontology, it will happen only after a great deal of
"data" (text chunks) is found in a particular area.  In other words, the
more chunks are placed in some part of the outline, the more conceptually
coherent the outline will become.  I think you can see this with my outline
of Hobbes' *Leviathan*: the "Metaphysics" part is skeletal and puzzling,
while the "Law" part is relatively well-developed.  Of course, that's just a
reflection of the fact that Hobbes himself discussed law a great deal more
than he did metaphysics, in *Leviathan*.

We didn't get to talk about project governance at all.  Next time, perhaps.

Let's do it again next week, same time; if not so many people show up next
week, maybe we'll make it every other week and/or try a different time.  But
I was happy with the showing, considering the relatively small number of
people on the Textop mailing lists (total of 40, maybe?).

By the way, if you want to participate, please do get your hands on a
headset.  It makes it easier for you and you reduce the chance that your
hardware set-up will cause noise that will force me to mute you.  :-)

--Larry