[seqfan] Re: Additional keywords for OEIS

[Antti Karttunen]

(Sorry, a bit late reply...)

On Tue, Feb 24, 2015 at 3:06 PM,  <seqfan-request at list.seqfan.eu> wrote:

> Message: 6
> Date: Mon, 23 Feb 2015 13:02:26 -0800
> From: Marc LeBrun <mlb at well.com>
> To: Sequence Fanatics Discussion list <seqfan at list.seqfan.eu>
> Subject: [seqfan] Re: Additional keywords for OEIS
> Message-ID: <D110D562.19B0B%mlb at well.com>
> Content-Type: text/plain;       charset="US-ASCII"
>
> It is my fond and enduring hope that the OEIS will someday enjoy the same
> level of sophistication in its data science content as it does in its
> mathematical content.
>
> Features such as ad hoc keywords or categories are innocuous insofar as they
> go, and useful within their scope, but they seem pretty primitive compared
> to the current state-of-the-art in data modeling and knowledge engineering.
>
> I recommend that anyone with an interest in these proposals with an eye
> towards the future of the OEIS should look into things like the semantic web
> to get a more up-to-date appreciation for where such technologies are going
> (indeed, have already gone).
>
> For example a keyword can be viewed as metadata that makes an assertion
> *about a sequence*.  But instead of having "?" variants of every keyword one
> can instead have a general kind of "meta" assertion *about an assertion*
> that says it is conjectural.  And so on for many other applications (eg
> assertions for provenance attribution, processing inferential consequences
> of keywords such as prim==>nonn, etc etc).
>
> The point is that there already exist sophisticated and elegant data
> languages for representing this kind of information.  The OEIS should
> consider adopting these approaches, rather than reinventing the wheel.
>
> But heck, even if we just go ahead and kludge stuff up the good news is that
> the total amount of actual data in the OEIS is truly miniscule by modern
> standards.  This relative footprint will continue to shrink, at least while
> the primary source of content continues to be entries hand-crafted by human
> beings.

On the contrary, I think that it is one of our best assets that our
entries are hand-crafted, and we are not (yet) polluted by all that
automatically generated "content" I see more and more in web.
Also, I wonder to which sources you compared our "relative footprint" ?

It should be noted also, that many of the entries are "dormant", in a
sense that their real significance will come apparent only after many
years. Although of course automatic tools like Weidmann's SequenceBoss
or any future Superseeker will have an important role in helping to
find out that significance.


>  This means that building and maintaining useful informational
> superstructures around the OEIS will lie well within the scope of
> contemporary tools and techniques, and they can flourish off-site if not on.
>
> With the current commercial interest around "big data" in many forms and its
> consequent stimulation of academia, maybe we could get, say, a coven of grad
> students to take a hands-on interest in this important aspect of the OEIS
> and help us cook up something more au courant?
>

Well, even before committing to any one schema in that field, it would
be a good idea to sort out keywords like listed in:

https://oeis.org/wiki/User:Charles_R_Greathouse_IV/Keywords/Table

according to whether they are such that software-bots can

1) only make conjectures about them (e.g. like currently, if the user
supplied terms do not contain any negative terms, then it is probably
"nonn", but without access to the real mathematical definition of the
sequence, or without ability to make such inferences as you pointed
out above, the server software or bots can never be sure),

or

2) make an instant affirmation (e.g. if the user supplied terms have
at least one negative term, then the sequence should surely be "sign",
i.e., "non-nonn")

or

3) keywords, like currently "base", whose meaning seems to stay vague
for time being. (Would k-automatic and k-regular be more exact
mathematical categories?)

And then, like you said, which keywords imply or exclude some other
keywords. And how to flag cases where a human being has disproved a
conjecture made by a bot, i.e., to distinguish conjectured and proved
categories.



Best,

Antti