===============================================================
Abstract (same as submission)
A Web Application for Finding Potential Collaborators
Marcia A. Derr and Karen E. Lochbaum
U S WEST Advanced Technologies
4001 Discovery Drive
Boulder, CO 80303 USA
{mderr,klochba}@advtech.uswest.com
One way to facilitate collaborations in an organization is to provide
tools for identifying potential collaborators. At U S WEST we have
designed and prototyped a tool for finding employees with particular
expertise. This work is part of a larger effort to provide technology
to support "knowledge sharing" within the company.
The expert-finding tool is implemented using two enabling
technologies, World Wide Web (WWW) and Latent Semantic Indexing (LSI)
[1]. WWW technology provides an easily accessible interface for
specifying queries and for presenting query results. It also provides
a means to link query results to other Web-based tools containing
related information. LSI provides a means of indexing and retrieving
documents based on semantic similarity. This is in contrast to other
full-text indexing techniques which require exact keyword matches.
The expert-finding tool operates in two stages: preprocessing and
querying. During the preprocessing stage, monthly accomplishment
reports written by first-level managers are automatically split into
"documents." Each document corresponds to a paragraph or bullet item
and describes an activity pursued by an employee during a particular
month; these documents provide the basis for the system's knowledge
about expertise within the company. The name of the manager who wrote
the report is prepended to each document to provide a point of contact
if no other name is included in the document. A large semantic space
(~100 dimensions) of document and term vectors is then created from
these documents using LSI.
In a second preprocessing step, a heuristic algorithm scans each
document to i) identify potential names, ii) look them up in an
employee directory accessible via a Web interface**, and iii) rewrite
each matched name as a hypertext link to relevant employee
information. The name of the manager who wrote the report is used toconstrain the search for other names within the report.
During the querying stage of operation, users access the
expert-finding tool via their favorite Web browser, and enter a query
using unrestricted English text, such as "Who has worked on knowledge
sharing and the World Wide Web?" The query is then submitted to LSI
which places the query in the semantic space of documents and returns
those documents that are closest to the query, as determined using a
cosine measure. The hyperlinks in the returned documents allow the
user to retrieve contact information, as provided by another Web
application, about the employees named in the documents. Other
hyperlinks allow the user to access the original source of the monthly
accomplishment reports to see the returned documents in context.
We believe that two factors will contribute to the usefulness of the
tool as compared to an earlier version deployed at Bellcore [2].
First, the availability and ease of use of Web browsers makes the tool
accessible to anyone with network access. Second, the tool's document
database is updated monthly, based on information already required by
the company for other purposes. This will ensure that the information
provided by the tool will not become stale.
FOOTNOTE
** We use "screen scraping" techniques to simulate sending a query to
a forms interface and intercepting the returned HTML. We would like
to see the development of Application Programming Interfaces (APIs) to
facilitate the interoperation of independent web applications.
REFERENCES
1. Deerwester, S., Dumais, S. T., Furnas, G. W., Landauer, T. K., and
Harshman, R. "Indexing by latent semantic analysis," Journal of
the American Society for Information Sciences, 1990, 41(6).
2. Streeter, L. A. and Lochbaum, K. E., "An Expert/Expert-Locating
System Based on Automatic Representation of Semantic Structure". In
Proceedings of the Fourth Conference on Artificial Intelligence
Applications. San Diego, CA. 1988.
===============================================================
http://www.rand.org/personal/hearn/collab.html
Using the Web for Dynamic Updating of Executing Processes
The REDUCE computer algebra community has been an active user of computer
networks for many years. The reason is obvious -- since contributors are
widely distributed across several countries, there is a need to build a
virtual community to achieve an effective collaboration.
This community quickly adapted to new capabilities made possible by
network advances. To supplement the standard distribution of REDUCE,
which occurs roughly every eighteen months,
gopher
and WWW servers were used
to facilitate code distribution to both developers and users.
Since REDUCE is developed and used worldwide on many different computing
platforms, portability has always been important. Source code portability
was achieved by defining a standard programming language based on Lisp
[1].
Later, a virtual machine [2] for this language was defined in terms of a set
of machine independent macros, which could then be expanded into assembly
language for a target machine, thus simplifying compilation.
However, different compiled code was required for each machine.
More recently, an ANSI C-based compiler [3] was
developed that produces
efficient machine-independent pseudocode as output. In other words, any
package in the system can now be compiled into this pseudocode, which can
then be executed by a running REDUCE process on any platform. Client
support comes in the form of an ANSI C-based interpreter for the
pseudocode, plus a function library to support standard operations such as
extended precision arithmetic, input-output and the like.
In spite of these advances, most system updating is still fairly
conventional. Developers send source code to a central repository, which
is mirrored in several countries . Users obtain updates from an appropriate
repository, compile this code on their client machine to build a
new executable image, and then run this image.
In this note we present a new model under development that utilizes the
characteristics of the World Wide Web for the dynamic updating of an
executing process. In this model, package developers maintain their own
repositories on a local server. A user starts a REDUCE job on a
client machine. If the calculation requires a package that has been
marked for possible updating, the executing process checks to see if a
more recent version is available on the package maintainer's server. If
it is, pseudocode for that package will be downloaded and used. If the
server is unresponsive, the existing package on the client machine is used.
Since executable code is being downloaded, there are obvious security
concerns. Initially, the user community will be a small group of trusted
contributors. However, in the long run, it will be necessary to face the
possibility of a virus being inserted in one of the executable packages.
We shall therefore study relevant models such as
HotJava to ensure the integrity of our
approach.
Our initial testing involves a group of approximately twenty active
developers and users. Once the model is stable, it will be offered to the
whole user community.
References
[1] A.C. Hearn, "REDUCE - A User Oriented Interactive System for Algebraic
Simplification," Interactive Systems for Experimental Applied Mathematics,
(edited by M. Klerer and J.Reinfelds), Academic Press, New York, 79-90,
1968.
[2] M.L. Griss and A.C. Hearn, "A Portable LISP Compiler,"
Software Practice and Experience 11, 541-605, 1981.
[3] A.C. Norman, "Compact Delivery Support for REDUCE," Journ. Symbolic
Computation 19, 133-143, 1995.
===============================================================
From: beethovn@ai.mit.edu (Ian S Eslick)
Abstract
The rapid growth in networked computing resources over the last decade
has drastically altered the computational landscape. No longer does
the majority of the world's computing resources reside in isolation,
but rather contribute to a global computational fabric. The World
Wide Web has pulled together this loosely-integrated, heterogeneous
collection of machines and provided a uniform way to publish and view
information in the network. While state-of-the-art Web technologies
provide many avenues for the evolution of human consumable
information, there remains little support for harnessing the software
and hardware computing resources underlying the Web. The Global
Cooperative Computing framework extends the Web by providing
unified mechanisms for viewing and managing many aspects of software
evolution. This includes development, distribution, execution,
feedback gathering, profiling and optimization. By allowing many of
these stages to progress automatically and transparently across the
Web, software is enabled to quickly adapt to new information and
changing requirements while delivering high-performance and
convenience to the end user.
===============================================================
http://web.cs.mun.ca/~lez/w3collab/abstract.html
Abstract for Participation in w3collab
Research into hypertext servers and browsers has been ongoing since
the mid 1980s in the context of electronic books (ebooks) within a
traditional publishing house. Several prototype browsers have been
developed independantly and in collaboration with others. A hypertext
language (Loose Hypertext -- LHT) and its derivatives have been
encorporated into a `hypertext compiler' which translates LHT
encoded ebooks into Web document structures.
A LHT ebook is analogous to a memex data construct which Vannevar
Bush called a `book on tape'. The LHT enhanced web server extends
the HTML hypertext model. Scope and default applied behaviors both at
compile time and at serve time relax some HTML restrictions. LHT
facilitates automatic link detection, view predicates based on user
class, bidirectional linking, default navigational topologies and
other flexible associations.
The LHT enhanced server is also capable of serving a rudimentary
trailblazing environment which allows thematic / semantic groupings to
be developed within a generic www browser. Trail following is
implemented using the `Electronic News Stand' metaphore. A
privelaged user can review and augment the hypertext in situ through a
standard web browser.
Development of the production service is being persued by a
consortium of mixed media producers in Newfoundland and
Labrador. Consortium partners represent traditional print, radio and
television producers and distributors. The system may also be
employed for telework and rural networking as the Enterprise Network
addresses its mandate to support community economic development
through the strategic application of information technology and human
mediation.
I hope to contribute to Web evolution in the following areas:
- Memex functions in the Web (eg. trails)
- collaborative language elements (non linear composition)
- advanced link concepts (direction, creation, types, implied, aggregates)
- group ware for publishers
- data topologies, distance and dimension
- commercial electronic books
- the Web for commercial publishers
- softening the protocol wrt copyright
- unusual data behaviours
August 31, 1995.
===============================================================
Proposal for a Presentation
WinWin: A System for Collaborative Generation
of System Requirements
by
Prof. Ellis Horowitz
Computer Science Dept.
University of Southern California
Los Angeles, Ca. 90089
horowitz@usc.edu
What is Winwin
WinWin is a system that supports the collaborative acquisition of
requirements for a proposed system. It is primarily aimed at
people who contract for, and develop software, but it is applicable
to establishing the requirements for a broad range of products and systems.
Of special interest is the WinWin model of collaboration, which
is based upon Theory-W principles. In simplified terms, collaborators
(called stakeholders) use Winwin to enter their Win Conditions for the
proposed system. Win Conditions must evolve to Agreements. However, Conflicts,
Issues or Uncertainties may arise. These are captured and various Options for
resolving them are proposed. A voting arrangement supports the negotiation
process for resolving disputes. The resulting database is a (densely)
interconnected graph of artifacts whose transitive closure leads towards the
system's requirements.
As a software system WinWin has several major components which include:
support for distributed interaction, database maintenance and control, and
multimedia attachments. WinWin is an X/Motif/Unix application currently
running on Sparc/SUNOS and HP/UX platforms. WinWin supports collaborative
acquisition of artifacts by using TCP/IP, datagram and socket technology.
IIt maintains the database of all artifacts, supports lockout and concurrency
control. Every artifact in WinWin can have attachments. These attachments are
described by a pair of pointers: (Software, File), where
Software is some executable program and File is a file that should be sent
to the software. Typical examples include associating sound and video
with specific artifacts. Currently WinWin can export its
database of artifacts to HTML. Links between artifacts are transformed into
hyperlinks.
WinWin Limitations with Respect to Web Browsers.
The export of WinWin data to HTML is a compromise decision,
necessitated by the limitations of HTTP and HTML, which prevent us
from fully implementing WinWin in these technologies.
In effect we currently have two parallel systems.
Both can be used to browse the hyperlinked database and to view the
multimedia attachments. Both support simultaneous viewing by geographically
distributed stakeholders.
But the Mosaic/Netscape browsers are insufficient to capture or store the
data. More significantly, the browser is not capable of imitating the
semantics needed when a vote on an agreement is initiated. In this case
all involved artifacts are frozen, except for comments. This is required
so that each stakeholder will view the same set of facts and vote on the
same conditions. Other aspects of WinWin that are either impossible or hard
to implement in HTTP are:
- stakeholders can only alter their own data, and not someone elses;
- stakeholders may point their data to someone elses;
- stakeholders share a common taxonomy, that is editable only by those
with permission;
- stakeholders may attach comments to someone elses data.
Suggested extensions to HTTP and HTML.
I have not thought this out in great detail. One fundamental extension
would be the power to invoke a variety of programs on the client side.
Using this ability I could build the WinWin data capture routines.
Alternately, if HTTP could support access to a dbms, then I could
use that to hold the database, which would substantially simplify
the WinWin implementation. Another issue is to identify and be aware
of the collaborators. During an active session, one person making changes
is required to notify others, and if they are simultaneously logged
on then the notification must appear on their screen.
Since the data in WinWin is tightly inter-related, but owned by different
stakeholder, votes must freeze the data, implying that all collaborators
mmust be immediately informed of any change of state.
=============================================================================
Ellis Horowitz ___ ___horowitz@pollux.usc.edu
Computer Science / / /__ / Fax: +1-213-740-7285
University of Southern California /__/ ___/ /__ Voice: +1-213-740-8056
Los Angeles, California 90089-0781 Sec'y: +1-213-740-4498
=============================================================================
===============================================================
From: dougross@mit.edu (Douglas T. Ross)
The following extract from my position paper: "Understanding /:\ The Key
to Software" -- prepared January 1989 (for the \Workshop on Complex
Software Systems\, in Austin, Texas, sponsored by the Computer Science
and Technology Board of the National Research Council) still is
pertinent to today's Working Conference theme. [See "Scaling Up: A
Research Agenda for Software Engineering" National Academy Press,
October 1989]
Briefly: "Understanding The Key to Software" + "Understanding: The Key
to Software" = "Understanding Understanding"!!! -- pretty general, you
have to admit!
I went to that workshop with a single-minded goal:
"To help to define a cohesive blending of our existing world-
wide communications networks with color-graphics hypertext advances so
that a \collegial community\...can interactively pursue joint endeavors.
... I propose that the resulting system be named "The Worknet", and that
the word "worknet" have verb and adjective, as well as noun, meanings
...[so that] people can worknet with others on worknet solutions as
active reader/author members of worknets [dedicated to specific topics].
... The point is that it is not the network that is important, nor even
the users -- it is their \collaborative work\ that matters.
"The Worknet architecture must be completely open, so that each
workgroup can supply or procure language features best suited to their
topic, without compromise. Each group will need both \document\
language and \comment\ language (perhaps the same) to be able to probe
the \meanings\ of documents, interactively through the network. The
single \universal design criterion\ that must not be violated [in the
provision of support services in implementing the architecture] is that
Comments ON a document do not affect
the form or meaning OF the document in any way.
With this distinction rigorously supported, the meaning of reader
comments ON form or meaning or both can reliably be incorporated into a
revision OF the document by its author -- for further circulation for
commentary. Every reader also is an author; every comment is a
document; and with meaning thus guaranteed to be preserved, true \team
understanding\ can be evolved and expressed for any subject of interest.
"The important feature not yet adequately supported (although it
could be with careful design and use of color) is to \graphically
portray\ the reverential meaning that is to be common between an
arbitrary comment-language expression and an arbitrary document-language
expression to which it refers -- with minimum disruption to either
document -- for understanding often derives from subtle features of the
\form\ or "look" of an expression in a language."
These notions came, of course, from the Reader/Author Cycle which always
has been a key component of SADT (Structured Analysis and Design
Technique) and its now-government-standard version IDEF0 (Integrated
DEFinition 0), FIPS#183, in the development and modeling of team
understanding of any subject of interest. I'm hoping to bring SA up to
date, make it useful on the Web, and to apply it in the communication of
W3C results to the many involved constituencies.
The delay in penning this note was due to my still-fruitless search for
some thoughts I wrote down recently on "consortium" vs. "collaboration",
and the like. The pertinent drift was that one "consorts" with friends
but may well "co-labor" with one's deadliest enemies, as well -- to the
extent that such joint work serves to further the interests of the
parties involved. Notice that both terms arise in today's Workshop!
I think it is increasingly important that we articulate why we use one
word rather than another so we build up a shared frame of mind in which
we can work together without misunderstanding -- as we focus on shared
\understanding\. (In SA, for example we name boxes and label arrows,
but NEVER mis-speak and label a box or name an arrow! It's \simply not
done!\ -- once the peer-group culture is absorbed.)
I mention this because the "co-laboratory" view of a shared and
distributed \laboratory\, which Bill Wulf (who also was at the Austin
Workshop, as he had just arrived at NSF, at the time) did so much to
promote during his stay at NSF, has these deeper etymological roots so
that "collaboration" can prosper in a still broader context -- in areas
where the idea of a "laboratory" might not fit, for some people. (Share
an intellectual hot-tub, anyone?)
===============================================================
Akihiro Kubota, Dr.Eng., Associate Professor
bota@race.u-tokyo.ac.jp
http://brains.race.u-tokyo.ac.jp/bota'sHome.html
RACE:Research into Artifacts, Center for Engineering
The University of Tokyo
4-6-1 Komaba, Meguro-ku, Tokyo 153, JAPAN
Tel:+81-3-5453-5885 Fax:+81-3-3467-0648
http://www.race.u-tokyo.ac.jp/
Curriculum Vitae
Dr. Akihiro Kubota has been Associate Professor at the Design Science
Division of RACE (Research into Artifacts, Center for Engineering),
the University of Tokyo, since 1992. He received his doctor's degree
from the Graduate School of the University of Tokyo in 1989.
Research Interest
Dr. Kubota is the project leader for the design and development of the
Abduction Machines. Abduction Machine (AbM) is a tool for supporting
global collaboration among designers/engineers over the World Wide
Web. He is also the co-leader of the "RACE" (RACE Asynchronous
Collaboration Environment) Project. The "RACE" Project aims at
developing methodologies of asynchronous collaboration using the
global network and media technologies. AbM is one of the key tools for
the "RACE" project.
The main target of AbM is innovative design aid. It is important to
foster creative environment where the team members from different
professions can submit various ideas freely, frankly and competitively.
Such collaborations are usually based on face-to-face communications.
However, since independent professional designers are often working in
distributed location, the design work should proceed without calling
face-to-face synchronous meeting frequently.
AbM is oriented to asynchronous collaboration environment. It mainly
supports sharing and version control of various design documents by
utilizing the World Wide Web. Since the collaborative design work is
basically multi-tasking, the synchronous collaboration which forces
single-tasking should be minimized. The asynchronous collaboration is
also easier to get across the difference of time or languages.
AbM is incorporating the following two sub-tools for rapid
prototyping, which is the key to innovative design.
Mental World Browser(MWB)
Physical World Transmitter(PWT)
MWB is a visualization tool of human mental world. MWB aids
prototyping of concept, which is called "abduction" (hypothesis
generation). The visualized mental world can be exported to shared
space by HTML/VRML files for the global collaboration.
PWT is a shape copy and transmission tool by using 3D laser scanner
and stereo laser lithography over VRML. It realize remote prototyping
(abduction of shape) using traditional clay models or sketches. These
tools increase the number of iterations of prototyping for designing
more elaborate products.
Since it is also important to promote local collaboration, AbM should
be a tool for fostering local creative environment. The external
design of AbM is now proceeding with the cooperation of Tokyo National
University of Fine Arts and Music. The external material of AbM is
sugi (Japanese cedar), which is deeply related to Japanese life and
culture from old times. AbM is composed of multiple parts. It permits
flexible reorganization according to the various human thinking modes.
-------
Regards,
Akihiro Kubota, Dr.Eng., Associate Professor
bota@race.u-tokyo.ac.jp
http://brains.race.u-tokyo.ac.jp/bota'sHome.html
----------------------------------------------------
RACE:Research into Artifacts, Center for Engineering
The University of Tokyo
4-6-1 Komaba, Meguro-ku, Tokyo 153, JAPAN
Tel:+81-3-5453-5885 Fax:+81-3-3467-0648
http://www.race.u-tokyo.ac.jp/
===============================================================