Hi Jack, (01)
Ok, I'll see if I can put something decent together over the Xmas
holidays to post up in the New Year. (02)
To speak to the 'langue' and 'parole' issue just briefly and sketchily,
because I am experimenting here too:
What I hope to be able to do one day is import a dictionary (ordinary
natural language) to provide a layer of atemporal 'ideas' (concepts).
What the structuralists said about 'langue' was that every term was just
defined in terms of other terms in that layer.
Concepts (words in the dictionary) are the class layer of language.
'Parole' was about the temporal usage of concepts in respect of the
world => instances.
My plan was to have a 'memory' engine that tracked what happened over
time as it scanned through texts to create a layer of memorised
instances and instance networks. Each word encountered in a text becomes
a bound instance of a concept. I.e. a subject that is/can be an
addressable resource at a time and that can be labelled with a name; the
name being the same string as the concept. This memory layer tracks the
state of actuality. (03)
The concept layer gives definition to the terms encountered in the
memory layer, the memory layer connects (proxy) instances to the world.
In Peircean terms the interpretant is missing in a way, because the
alignment of instances to ideas just takes place through string matching.
Possibly some of that is just a function of only needing string matching
to align the layers at that point, and perhaps some is down to not
really having any agent goals in respect of the perception of
world-things/actuality... because I haven't got anywhere near building
anything like that yet. (04)
What I meant about the backlinks w.r.t. to tagging, was that the names
attached to the links are not embedded in any material text resource.
There is no way to traverse back up to a surrounding text that provides
some context to the attachment of the label(s) to the link. The
surrounding 'text' is all in the tagger's head. (05)
Atb,
Peter (06)
Jack Park wrote:
> Peter, this is really dense and thoughtful, certainly worthy of
> expanded consideration. I have a few comments interspersed, but I'd
> love to read an expanded paper with diagrams (I think in diagrams)
> that give examples of what you are saying. I'd like to see it related
> back to Mark's original inquiry.
>
> Many thanks
> Jack
>
> On 12/22/05, Peter P. Jones <ppj@concept67.net> wrote:
>
>>
>>Jack Park wrote:
>>
>>>Mark, what you said:
>>>"..in collaborative work you have multiple intersecting
>>>topic spaces (our understandings of what "topic space" is may not be
>>>the same). I need code assist to manage and tag my topic spaces, and
>>>then we need further code to assist in managing the intersections of
>>>our topic spaces."
>>>drives my intuition engine into overdrive; I'll need to double my dose
>>>of adderall in order to work on that.
>>>
>>>My friend Bernard Vatant started a weblog devoted to all things
>>>related to subject identity [1]. The notion that our understandings of
>>>what "topic space" is may not be the same, it seems to me, seeds the
>>>need to recursively waltz deeper; just what is a topic in the first
>>>place? The TMRM specification has always stated that a subject is
>>>whatever you want to talk about, and that a "topic", in a physical
>>>sense, is a proxy for that subject.
>>>
>>>I have been evolving this notion that one of the use cases for subject
>>>maps (topic maps according to the TMRM) is that of culture fusion. I
>>>am looking for ways to articulate the intuition that subject maps
>>>allow for the representation and expression of diverse world views; I
>>>think the term "world views" speaks to what you mean by context we
>>>share. I have heard Steven Newcomb say that, to do this, one needs to
>>>know that the subject being discussed is the same subject for all
>>>world views under discussion. Without that, it would seem, the
>>>opportunity to federate without subjugation will be difficult.
>>
>>Perhaps...
>>Imagine an engine which separates the 'langue' from the 'parole', or put
>>another way, the world model from the memories of usage. Perhaps then
>>what becomes clearer is the way that strings can be identical but the
>>bindings to the world can be unique in some degree.
>>In order to ascertain that two things are the same subject it would seem
>>to be necessary to know the nature of the binding in detail, rather than
>>the label the subject is given.
>
>
> What I really want to understand is the separation of 'langue' from
> the 'parole'.
>
>
>>This always seemed to me to be the core of the distinction between topic
>>name and the subject identifier, and the reason subject identifiers were
>>chosen as a separate basis for merger with equal primacy with name strings.
>>One of the reasons I preferred the ISO 13250-1999 draft and my OTM
>>approach to later TM merger rules was because the possible absence of
>>merger process allowed the choice of views (the points of mergers)
>>dynamically.
>>That is, I could ask of the system, 'what topics have the same binding'
>>or 'what topics have the same name' and be comfortable with the
>>difference in the answers.
>
>
>
> As I have written TopicSpaces, you can ask for subjects based on
> bindings or on names.
>
>
>>Topics with subject identifiers had bindings to bits of the world.
>>Associations between topics without bindings are just concept maps.
>>The difference between the two is similar to the difference between the
>>class layer of concepts and the bound layer of instances.
>
>
> Let's back up here and ponder what "subject identifiers" means. There
> are all sorts of subject identifiers, including the PSI (a kind of URI
> artifact created either by fiat or concensus), and clusters of subject
> identity properties which allow to disambiguate subjects. More
> recently, there is the notion of "conferred subject identity", that
> which can occur in various ways. "Topics with subject identifiers had
> bindings to bits of the world" I'd like to see examples of that.
>
>
>>The bindings of instances take place in contexts.
>>Sometimes common usage of a conceptual name has a pre-defined context
>>(as in the technical use of a word in, say, mathematics as opposed to
>>general use). But its context is composed of other concepts, not
>>instances with bindings to the world.
>>Whereas the binding of an instance takes place in a context that can
>>contain, or be composed of, other instances.
>>Also, the conceptual does not really track state changes very much. It
>>resists change. Whereas the instance layer is much more like memory,
>>where state changes in actuality are logged.
>>
>
>
> Like to see concrete examples of that...
>
>
>>Intersections of topic spaces are more about the memory layer, it seems
>>to me and federation might not be complicated if you can let the machine
>>happily ignore all notions of worldview in that layer.
>>
>
>
> Here is an opportunity to tie "memory layer" back more concretely to
> what was said above. I'm hoping for examples.
>
>
>>That might be the case for one natural language - say US English, but
>>other issues arise from considering translations with other languages.
>>But these are more concerned with mappings between the concept layers.
>>
>
>
> The "concept layers", the "memory layer". Need a picture :)
>
>
>>Under this model (analogy?), link tags are the attachment of descriptive
>> names to a link (a set of bindings). I.e. each of those tag names
>>acquires an instance binding to the resource. The resource acquires a
>>set of names. But sadly, most of the context of the binding act(s) gets
>>lost in tagging aggregation sites, because there are no useful backlinks.
>>
>
>
> Backlinks to what? To some imagined context? Let's play with this.
> Some imagined tagging aggregation site has it that some imagined user
> sees some resource, somehow imports a reference to said resource to
> the tag aggregation site, and "bookmarks" it with a tag. The presumed
> benefits include having ones bookmarks online rather than stuck on
> some platform at home, work, etc. Also, there is this social discovery
> process whereby people get to see what others also have seen by
> following links around the graph that includes resources, tags,
> bookmarks, and people. Backlinks to what? To the context in which some
> user created some tag? How would one create such a backlink if that's
> what is meant by "no useful backlinks"? Could the tag artifact include
> some commentary, perhaps even links to contextual resources the author
> of the tag thought important?
>
>>--
>>Peter
>>
>>
>>>Waltzing deeper, one wonders just what role tag clouds bring to the
>>>table. In subject maps terms, they are "names for things". What
>>>things? Rather concrete things, in the case of bookmarks. Tags serve
>>>as binding points, precisely the notion of what a topic exists to
>>>provide.
>>>
>>>I wish I had more time right now to stretch this, but I'm very close
>>>to bringing up a tagging engine sitting on top of a subject map. I can
>>>hear my jvm calling me..
>>>
>>>Cheers,
>>>Jack
>>>[1] http://universimmedia.blogspot.com/
>>>
>>>On 12/21/05, Mark Szpakowski <szpak@well.com> wrote:
>>>
>>>
>>>>Regarding metaphor, I think analogy is actually a more fundamental
>>>>concept, with metaphor being a concretization of a bridging that has
>>>>already happened through analogy. The interesting question is how
>>>>analogy, or analogizing, serves to create new meaning, or new
>>>>abstractions. If the way you use the term X relative to [a-1..a-m] is
>>>>analogous to (= has some mapping to?) how I use the term Y relative
>>>>to [b-1..b-n], and if a's and b's are roughly the same, then do X and
>>>>Y refer to the same subject? And does binding X and Y under the same
>>>>subject release more information? This is where Victoria Livschitz's
>>>>"reverse inheritance -- the ability to define a new entity as a
>>>>generalization of an existing one" seems relevant - although "of
>>>>existing ones" might be more how it is. Human language seems to be
>>>>able to do this, or rather humans using language seem to be able to
>>>>do this. With greater ability to let our language loose in common
>>>>associational spaces, as seems to be happening with web pages which
>>>>link to one another, and with pools of folksonomically tagged pages,
>>>>there's some feeling of wanting to extract, extrude, generate,
>>>>further meaning. Note I use metaphor as I grope. The practice of
>>>>displaying visual tag clouds with tag labels sized proportionally to
>>>>their use-frequency is one attempt to extract meaning through metaphor.
>>>>
>>>>Needless to say, in collaborative work you have multiple intersecting
>>>>topic spaces (our understandings of what "topic space" is may not be
>>>>the same). I need code assist to manage and tag my topic spaces, and
>>>>then we need further code to assist in managing the intersections of
>>>>our topic spaces. I automatically, or inevitably, inject my contexts
>>>>onto your words. Sometimes I feel we share contexts. Sometime I feel
>>>>we need to delineate our different contexts, and that that's
>>>>fruitful. All of this is our workspace.
>>>>
>>>>Cheers on this solstice day,
>>>>Mark
>>>>
>>>>
>>>>On Dec 21, 2005, at 12:29 AM, Jack Park wrote:
>>>>
>>>>
>>>>
>>>>>Yup. Hibernate. It's what's for dinner, sometimes. I spend a lot of
>>>>>time trying to decipher the many ways their interpreter will handle
>>>>>what I toss at it; still trying to get it to honor my requests for
>>>>>primary key. It works, primary key or not.
>>>>>
>>>>>What I was thinking had to do with some form of declarative thinking
>>>>>that would build subject maps. That can take many forms, though I
>>>>>haven't wrapped my head around metaphors yet; far too busy bringing up
>>>>>a social application with core TopicSpaces, mostly trying to sort out
>>>>>a high level way to create assertions, since that's where the real
>>>>>power lies, asserting things.
>>>>>
>>>>>Dependency injection of context into a subject map. I need to think
>>>>>about that one. Here's a sketch of context as I see it in TopicSpaces.
>>>>>According to subject map terminology, one needs to disclose how it
>>>>>works, how you test for subject identity sameness, etc. My
>>>>>interpretation of that is that you build an engine that crafts a
>>>>>framework on which you can layer applications, each application
>>>>>forming a context. Consider wrapping my engine with, say, a social
>>>>>bookmarking application. People who go to that URL have only as their
>>>>>context that which is associated with bookmarks: bookmarked resources,
>>>>>people, tags and other comments. If that is all going on over the top
>>>>>of a given subject map, other applications may be using many of the
>>>>>same subject proxies for other things, but those other things are not
>>>>>presented in the context of any other application. That's just one
>>>>>view of context. It's unclear to me where metaphors will play out
>>>>>here, but I can't say I've had time to wrap my head around it. Maybe
>>>>>you've got some ideas, Mark?
>>>>>Thanks
>>>>>Jack
>>>>>
>>>>>On 12/20/05, Mark Szpakowski <szpak@well.com> wrote:
>>>>>
>>>>>
>>>>>
>>>>>>On Dec 20, 2005, at 1:04 AM, Jack Park wrote:
>>>>>>
>>>>>>Along the same lines, I recently gave a talk in which I speculated on
>>>>>>the benefits to splicing conceptual graphs into subject maps as
>>>>>>one of
>>>>>>the assertion patterns. No matter what we do at that level, I sense
>>>>>>that there would be great benefit to thinking through the ways in
>>>>>>which we craft our code. Recent playing with hibernate suggests there
>>>>>>are some other approaches.
>>>>>>Are you referring to Hibernate object-relational mapping through
>>>>>>an XML
>>>>>>file?
>>>>>>Do you see that as a model for a kind of dependency-injection of
>>>>>>context
>>>>>>into
>>>>>>subject maps?
>>>>>>
>>>>>>
>>>>>
>>>>>--
>>>>>This message is archived at:
>>>>>
>>>>>http://collab.blueoxen.net/forums/cgi-bin/mesg.cgi?
>>>>>a=yak&iQ79aafa0512202029p5914f53fo7afe57838a639a7@mail.gmail.com
>>>>>
>>>>
>>>>--
>>>>This message is archived at:
>>>>
>>>>http://collab.blueoxen.net/forums/cgi-bin/mesg.cgi?a=yak&i=4202BAAD-B193-4274-99B0-4364FE0C0DF1@well.com
>>>>
>>>>
>>>
>>>
>>--
>>This message is archived at:
>>
>>http://collab.blueoxen.net/forums/cgi-bin/mesg.cgi?a=yak&i=43AA930A.3010103@concept67.net
>>
>>
>
> (07)
--
This message is archived at: (08)
http://collab.blueoxen.net/forums/cgi-bin/mesg.cgi?a=yak&i=43AAEC90.8020107@concept67.net (09)
|