cjoslyn@bingvaxu.cc.binghamton.edu (Cliff Joslyn) (01/06/90)
I have in mind something very similar to what is being discussed here.
The general need is obvious, although we all have our special purposes.
The following is mine. I'm about ready to kludge something together
with email, TeX macros, and shell scripts, but perhaps we can all come
up with something better.
Contact me for more information.
_THE NECESSITY OF A NEW TOOL FOR PHILOSOPHICAL DEVELOPMENT_
Cliff Joslyn
Systems Science
SUNY Binghamton
Copyright 1990 Cliff Joslyn
Past development of philosophical texts and philosophical systems
rested on existing technology: paper and ink. This medium was
conducive to the construction of linear documents, written and read
from the begining to the end. While the size of these linear forms
could be greater or lesser, from long treatises to short paragraphs or
sections [e.g. Aristotle 1943, Wittgenstein 1953], over the years this
linear form has been maintained.
In recent years dictionaries, encyclopedias, and other reference works
have been developed which partially introduce non-linear structures
through internal references [e.g. Edwards ed. 1967, Krippendorf 1986,
Flew 1979, Raymond and Topa 1988, Kluwer 1988, Wiley 1979]. Some have
made halting efforts in the direction of non-linear documents [e.g.
Minsky 1986], and others have used pictures to aid in understanding
[von Foerster 1971, Abraham and Shaw 1985]. And certainly formal
systems (mathematics and logical notations) have given the ability to
construct large, complex linguistic systems.
But to date, individual scholars working in natural language have been
essentially limitted to traditional development forms. The development
of large, complex systems of philosophical thought in non-formal
domains have been difficult, and collaborative work on such systems
next to impossible.
Scholars who wish to develop such systems of thought require a new
tool. In particular, the Principia Cybernetica project [Joslyn and
Turchin 1990, Turchin and Joslyn 1989] demmands the use, and possibly
the development, of a computer-based tool to provide a new form of
linguistic expression. While this project will use such a tool for
specifically philosophical development, a tool is envisioned that could
be used to support scholarly and scientific writing, research, and
development in general.
CRITERIA OF THE PRINCIPIA CYBERNETIC PROJECT
The following is a partial list of desiderata for the "Principia
Cybernetica" project:
1. For a group of researchers, perhaps not all geographically close,
to collaboratively develop a system of philosophy, where
philosophy is taken in the general sense of clear and consistent
language about ideas and concepts;
2. To allow these researchers different levels of access to the
system according to their role in the project development;
3. To produce a system of philosophy that can develop dynamically
over time, with continuing refinement and expansion;
4. For the system of philosophy to fully reflect and incorporate the
semantic relations inherent among the terms being explicated;
5. To allow the explication of terms and senses of terms, and to
unify and synthesize notations and terminology among researchers
in different disciplines;
6. To support the process of argument and dialog among experts toward
the end of consensus at the level of the meanings of words and
the relations among those meanings;
7. To support the publication of intermediate and final stages of
parts or the whole of the philosophical system;
8. To support bibliographical and historical reference;
9. To support mathematical notation and the easy movement among
natural language, formal language, and mathematics;
10. To allow researchers to develop or read the philosophical system
in various orders and in various degrees of depth or specificity;
11. To allow access to the system for both participants who wish to
author and users who wish to read, browse, or study;
12. To support the publication of various special-purpose documents,
including dictionaries, encyclopedias, texts on a subject,
reference pages, essays, dialogs on a subject, or "streams of
consciuosness";
13. To allow the representation and utilization of knowledge in both
its breadth and its depth.
ON FORM AND CONTENT
There is a mutual affect between the form of an expression (the medium,
or syntactic language) and its content (the message, or semantic
meaning): the existence of linguistic forms constrains and allows the
possible expressions; and simultaneously the desire to express
additional meanings demands the changing of the language.
The form and content of linguistic systems evolve together, and such a
tool will evolve with the system of philosophy itself. A complete
implementation will not be possible before beginning development of the
philosophy. However, an ideal final state for such a tool can be
envisioned at this time, and a minimal implementation with existing
technology is possible with relatively little effort.
THE NATURE OF SUCH A TOOL
_Logical Structure_
Define a semantic network S = ( N, R ) [Rich 1983 pp. 213, Shastri
1988, Evens ed. 1988, Brachman 1977]. Define a set of nodes N = {
ni }, each of a specific node type. Node types would certainly
include traditional document units like words, sentences, and
paragraphs. But also included would be special document units like
bibliographical references and equations, and other types like
pictures, moving pictures, sounds. Nodes can be complex, that is any
ni can be composed of further nij, for example a paragraph composed
of a number of sentences and a picture.
Further define a set of relations R = { Rj }, where each Rj is in
the full cross product of all the ni. When ni1 Rj ni2 we say
that there is a link of type j from node i1 to node i2. The Rj
represent semantic relations among the nodes. For example, one
relation R1 could be that of definitions, where n1 R1 n2
indicates that the term represented by node n1 is defined in node
n2. Then the inverse relation R1-1 would indicate those nodes
referencing that definition.
An example would be where the term _entropy_ is defined in using the
terms _probability_ and _summation_; or, where _Shannon entropy_ is defined
using _entropy_. But at the same time, a link to _thermodynamic entropy_
and _irreversible processes_ would be established, thus maintaining full
semantic linkage throughout the system.
Other examples of semantic relations could include R2 being simple
reference, where n1 R2 n2 indicates that n1 is referenced in
the node n2. Other examples include implication and/or
generalization (e.g. probability to Dempster-Shafer theory, Newtonian
to relativistic equations of motion), historical precedence,
bibliographical reference, alphabetical order, author groupings,
etc.
_Complexity of the Semantic System_
One point that deserves consideration is the overall complexity of the
network and the relative number of nodes and relations, |N| and |R|.
One criticism that is made of semantic networks is the necessity to
maintain a large number of types of links in order to reflect the
semantic richness of natural language and philosophical concepts. It
is claimed that this makes semantic networks confusing to implement and
manage, because of the great number of links involved, allowing authors
and browsers to "get lost in hyperspace" [e.g. Brockman 1989].
However, the maximum number of links in our system S is:
|R| ( |N|2 - |N| ),
which grows faster with the number of nodes N than with the number of
link types R. Further, it has recently been recognized [e.g. Kanerva
1988] that complex systems are characterized by a relatively high
dimensionality |N| and relatively low cardinality |R|. This makes such
semantic nets systems in that class, and makes the repertoire of tools
developed for the management of complex systems available for this
project. But in any event, it will be certainly be necessary to
carefully maintain an appropriate balance between these quantities |R|
and |N|, and a relatively sparsely populated network to preserve
tractibility.
_Uses of the Relations_
Since each of the relations Rj represents a distinct semantic
category, they can be very useful for accomplishing specific ends while
using the network. For example, when the Rj are partial orderings,
then it is possible to recognize hierarchical relations among the
nodes. This would be the case, for example, for the relations of
definition or entailment. This would allow users and authors to ascend
manage. Linear orderings may also be available, for example
alphabetical or chronological order.
When the whole network is projected through one or more of the various
relations, it is possible to view the network or produce specific
documents of a specific form. In this way, semantic networks can be
used to produce the following kinds of special systems:
1) Encyclopedias (alphabetical order by subject, deep content)
2) Dictionaries (alphabetical order by subject, shallow content)
3) Bibliographies (alphabetical order by author)
4) Histories (chronological order)
5) Reference notebooks (subject order, shallow content)
6) Textbooks (subject order, deep content)
7) Subject-oriented expositions (ad-hoc order)
8) Development of arguments on specific themes (ad-hoc order)
_Existing Similar Work_
Reference works in Systems and Cybernetics currently exist [Singh 1987,
Krippendorff 1986, Ashby 1965] and others [Don McNeil, Len Troncale,
Carl Slawski] have approached this issue in the past. The whole or
portions of these existing works can be extended into semantic network
structures, as can or have other existing reference works [e.g. Flew
1979, Raymond and Topa 1988, Kluwer 1988, Wiley 1979]. We will
actively pursue the cooperation of these authors and editors.
IMPLEMENTING THE TOOL
Beyond considering the logical design of this tool, there is the task
of implementation.
_Necessary Features_
It will be necessary for any implementation of the tool to include the
following:
1) _Electronic Mail_
Since participants in the project live in different places, it will
be necessary for them to use telecommunications to carry out
development. This includes the use of one or more of the
international computer-based telecommunications networks (e.g.
BITNET, INTERNET, USENET). Portions of the semantic network can
be developed by individuals at one site and them distributed to
the others, or to a central site, using email. The actual
operation of the network should be available on a remote-control
basis by users at other sites.
2) _Portable Implementation and Hybrid Environments_
The participants in the project will also be based on different
computer systems using different operating systems (e.g. UNIX,
VMS, CMS, DOS) and display and printer devices. Thus it will be
necessary for the tool to work in these different hardware and
portability and adherence to industry standards will be
critical. Some participants may have trouble accessing
computer systems at all, and so at least a rudimentary mixed
electronic/print capability is needed.
3) _TeX Output_
Since it is intended to be able to submit either a projection of
the network or the whole of the network at any time for
publication, it will be necessary to produce documents that
publishers will accept. It will further be necessary to be able
to format the resulting documents in a variety of ways, for local,
high quality printing at participant sites or for electronic
submission to publishers. The current industry standard for these
capabilities is the TeX electronic typesetting language [Knuth
1979]. TeX files use a plain ASCII format, and TeX is implemented
for a great many operating environments. This allows maximum data
portability across environments. TeX also provides full support
for the formatting of mathematical equations and, using the BibTeX
extension, of bibliographies.
4) _Version Control_
As nodes in the network are developed by multiple authors, it will
be necessary to track their change [see also Irish and Trigg
1989]. This will allow contributors to see the changes made by
others, and the recovery of previous work. Version control can
either be done on the basis of maintaining a history of the nodes
in the form of differences between various versions (as is common
today for maintaining software source code), or in terms of a
"markup language" [see ISO 1986] which maintains editing and
revision marks in the body of a text file.
_Currently Available Software_
Of course, the desirability of such tools to support the construction
of knowledge systems and scholarly activity has been seen for many
decades. Beginning in the 1960's articles and books have been written
and software systems have been developed to fill these needs. Today
there are a great many tools and systems available [e.g. Kimura 1986,
Seiler 1989]. Also, other people are pursuing projects which are quite
similar to the system envisioned. For example, the Matrix of
Biological Knowledge is an effort to create a distributed, coherent
knowledge system for the biological sciences [see _The Matrix of _
_Biological Knowledge Newsletter_, edited by Dan Davison, Theoretical
Biology, T-10 MS K-719, Los Alamos National Laboratory, Los Alamos, NM
87545].
But by far the technology with the greatest applicability to the
current problem is that of Hypertext and Hypermedia [e.g. Smith and
Weiss 1988, Kahn 1988, Kahn and Meyrowitz 1988, Shneiderman 1989,
Conklin 1987, Louie 1989, and see the journals _Hyperage_ and _Hypermedia_]
originated by Ted Nelson [1965]. A true Hypermedia system is very
close to the ideal logical form described above.
Thus the first step in implementing this tool is the investigation of
existing Hypermedia implementations. That task has just begun, but the
following is a list of commercially available systems:
Company Name System Name
----------- -----------
OWL Guide
Xanadu Xanadu
Brainpower ArchiText
Software Aristry Hypertext Programming Library
Texas Instruments HyperTRANS
Neil Larson Houdini
Symantec Grandview
MetaGeneric Hypertext Viewer
Scribe KMS
MCC Planetext
Apple Computer Hypercard
Xerox NoteCards
Brown University Intermedia
Cognetics HyperTIES
SNeps
Each of these products has different strengths and weaknesses,
satisfying to varying degrees the necessary conditions above. To the
extent that I have investigated them, none seems ideal or even
sufficient. And all of them cost money.
_Minimal Immediate Implementation_
It would, however, be possible to implement a preliminary (and
rudimentary) form of the tool at the present time. The basis for such
an implementation would be the UNIX operating system available at the
SUNY Binghamton computer center. Nodes would exist as individual text
files written in TeX format. Such files would be fully available to
electronic mail, and version control is available through the RCS
Revision Control System.
Custom software would be written to maintain the relations among the
nodes. The software would be written in ANSI C for maximum
portability, mixed with UNIX system software for text manipulation
(e.g. the 'grep' and 'awk' programs). Ideally, links would be
indicated by special macros written in TeX and then managed by the
custom software.
Implementation of a full-fledged hypertext development environment
would be a very large task, so such an implementation is seen as a
temporary stage, until a sufficiently powerful tool was available
commercially and there was a budget available to afford it. The extent
of the customer software would depend mostly on the time available for
design and implementation. It will probably require a fair amount of
time to operate the software, including formatting of documents and
manual management of version control and operation of the software
system. However, the implementation must be sufficiently general to
allow whatever structure exists to be importable by the commercial
product once it is available.
REFERENCES
Abraham, Ralph, and Shaw, Chris: (1985) /Dynamics: the Geometry of
Behavior/, v. I-III, Ariel Press
Aristotle, : (1943) /On Man in the Universe/, Walter J. Black Inc.,
Roslyn NY
Ashby, Ross: (1956) /Introduction to Cybernetics/, Methuen, London
Barrett, Edward: ed. (1989) /Soc. of Text: Hypertext Hypermedia + Soc.
Const.of Info/, MIT Press, Cambridge
Brachman, Ron J: (1977) "What's in a Concept: Struc. Found. for
Semantic Nets", /Int. J. Man-Machine Studies/, v. 9, pp. 127-152
Brockman, R John, and Horton, W et. al.: (1989) "From Database to
Hypertext via Electronic Publishing", in: /Society of Text/, ed.
Edward Barret, pp. 162-206, MIT Press, Cambridge
Conklin, Jeff: (1987) "Hypertext: An Introduction and Survey", /IEEE
Computer/, v. 20:9, pp. 465-472
Edwards, Paul: ed. (1967) /Encyclopedia of Philosophy/, v. 1-8,
MacMillan, New York
Evens, Martha W: ed. (1988) /Relational Models of the Lexicon/,
Cambridge U, Cambridge
Flew, Antony: ed. (1979) /Dictionary of Philosophy/, Pan Books,
London
Irish, Peggy M, and Trigg, Randall H: (1989) "Supporting Collaboration
in Hypermedia", in: /Society of Text/, ed. Edward Barret, pp. 90-106,
MIT Press, Cambridge
ISO: (1986) /Text + Office Sys.: Standard Gen. Markup Language(SGML)/
Joslyn, Cliff: (1990) "Necessity of a New Tool for Philosophical
Development", NOTE: To be published
Joslyn, Cliff, and Turchin, Valentin: (1990) "Introduction to the
Principia Cybernetic Project"
Kahn, Paul: (1988) /Linking Together Books/, in: /IRIS TR # 88-8/
Kahn, Paul, and Meyrowitz, Norman: (1988) /Guide, HyperCard, and
Intermedia/, in: /IRIS TR # 88-7/, NOTE: Inst. for Res. in Inf. +
Schol
Kanerva, Penti: (1988) /Sparse Distributed Memory/, MIT Press,
Cambridge
Kimura, GD: (1986) "Structure Editor for Abstract Document Objects",
/IEEE Trans. Software Engineering/, v. 12:3, pp. 417-435
Kluwer Publications: (1988) /Encyclopedia of Mathematics/, Hingham MA
Knuth, Donald: (1979) /TeX and Metafont: New Directions in
Typesetting/, Digital Press, Bedford MA
Krippendorf, Klaus: (1986) /Dictionary of Cybernetics/, NOTE:
Available through ASC
Louie, S, and Rubeck, F: (May 89) "Hypertext Publishing and
theRevitalization of Knowledge", /Academic Computing/, pp. 20-23,
NOTE: SS lib
Minsky, Marvin: (1986) /Society of Mind/, Simon and Schuster, New
York
Nelson, Theodor H: (1965) "The Hypertext", in: /Proc. Int.
Documentation Fed. Conf./
Raymond, Darrell R, and Tompa, Frank W: (1988) "Hypertext and the New
Oxford Dictionary", /Communications of the ACM/, v. 31:7, pp. 871-879,
Rich, Elaine: (1983) /Artificial Intelligence/, McGraw-Hill, New York,
Seiler, Lauren: (1989) "Future of the Scholarly Journal", /Academic
Computing/, v. September, pp. 14
Shastri, Lokendra: (1988) /Semantic Networks/, Morgan Kaufman, Los
Angeles
Shneiderman, Ben: (1989) "Reflections on Authoring, Editing + Managing
Hypertext", in: /Society of Text/, ed. Edward Barret, pp. 115-131, MIT
Press, Cambridge
Singh, MG: ed. (1987) /Systems and Control Encyclopedia/, Pergamon,
Oxford
Smith, John B, and Weiss, Stephen F: (1988) "Overview of Hypertext",
/Communications of the ACM/, v. 31:7, pp. 816-819
Turchin, Valentin, and Joslyn, Cliff: (1989) "Cybernetic Manifesto",
to be published
von Foerster, Heinz: (1979) /Cybernetics of Cybernetics/, ed. K.
Krippendorf, Gordon+Breach, New York
Wiley Publications: (1979) /Citation Indexes/, New York
Wittgenstein, Ludwig: (1958) /Philosophical Investigations/,
MacMillan, New York
--
O------------------------------------------------------------------------->
| Cliff Joslyn, Cybernetician at Large, cjoslyn@bingvaxu.cc.binghamton.edu
| Systems Science, SUNY Binghamton, Binghamton NY 13901, USA
V All the world is biscuit shaped. . .UH2@PSUVM.BITNET (Lee Sailer) (01/06/90)
I'd like to thank Cliff for his long exposition, and make a few comments.
The idea of HyperText is very appealing, but as Cliff says, its implementation
on the grand scale will be long and difficult. it seems unlikely to me
that any attempt to analyze, design, and construct such a system will fail.
At the same time, such systems will evolve, borrowing ideas here and there till
such capabilities are widely available. By analogy, I offer our experience
with electronic mail, where despite numerous attempts to offer its capabilities
to the general public, it is finally something like FAX that really takes
off.
My second point is that a general purpose distributed semantic network
architecture is not the hard part. The hard part is deciding what to
put in the general purpose distributed semantic network. To make these
decisions people will have to debate endlessly on the merits of this or that
approach. That's the point of the previous groupware thread---how can we
create an efficient discussion and decision making environment?
It seems to me that there are two tracks evolving. On is to take a
system like email or UseNet and try to add more structure to it, and the
other is to take a system like HyperText and try to make it more dynamic
so that is can manage discussions and debates as well as archival knowledge.
Fun, eh?
lee