This is a request for a grant of $160,000 for two years of
research into computer science  and communication science
problems connected with the development
of an electronic newspaper.

	The Stanford Electronic Times would offer the following services
to its users through terminals with keyboards and CRT displays:

	1. The ability to search the current wire service news for stories
on topics of interest.  News for a longer period (up to a year) will also
be stored and can be searched.

	2. Local information some of which changes every day like campus
news and some of which changes more slowly like movie and restaurant
guides.

	3. The ability to give an interest profile and have stories relevant
to this profile called to his attention.

	An editor will create summaries and an \F1information thread\F0
through the news structure.

	The following kinds of computer science and communication science
problems arise in developing this service:

	1. Man-machine interaction.  Many of the users of the system will
be casual and many more will be occasional.  The casual user should be able
to get some use from the system without reading anything longer than a
placard on the terminal.  The occasional user should be able to use the
system after a long absence without having to relearn it.  This requires the
development of a uniform and easily remembered style of interaction.
The experienced user should be able to get what he wants with very brief
commands.  We have many ideas we would like to try about what kind of style
will satisfy all these kinds of users reasonably well.

	2. File handling and retrieval techniques.  The Associated Press
A-wire puts out about 75,000 characters per day.  An index to this news
At present we build a hash table structure that allows immediate access
to all stories containing any non-trivial word.  This structure requires
an amount of memory that can reasonably be kept in core memory.  When we
go to larger amounts of data by adding more sources of news and allowing
the search to range over longer time spans, new data indexing and
handling techniques will have to be developed.

	3. We hope to develop good ways of describing the interest profiles
of users and automatically finding stories conforming to the profile.

	4. The system will involve at least two computers, the PDP-10
time sharing system at the Stanford Artificial Intelligence Laboratory
and the IBM 360/67 operated by the Stanford Center for Information
Processing.  This is because users of both systems will want access to
the newspaper and neither computer can afford to handle the whole load.
Besides this, some other computer centers on the ARPA network are
interested in the service and can also subscribe to the wire services
which will reduce the loss of information resulting from computer down
time.  Making these disparate computers interface smoothly and divide
the work is an interesting problem in computer science that will have
application to other information systems.  This is because the long
range goal of making all the information in the country accessible
from any place in the country is not being solved by creating a single
grand national information system.  Instead many information systems are
being developed with limited goals and will need to be interlinked.

(The development of a system for describing pre-existing files and and
the development of a system that will answer questions that require
calling up other computers and using their pre-existing files will
be the subject of another proposal).

	5. The electronic newspaper provides new scope for journalists.
The strict space limitations of newspapers disappear which permits
information to be presented in greater depth.  On the other hand,
techniques have to be provided form permitting the user to comfortably
find what he wants in a larger body of information than is contained
in a printed newspaper.

	Much better facilities for reader response can be provided.
Most likely, all non-libelous letters to the editor can be made
available, and facilities can be provided for permitting a reader
to see what responses have been made to articles expressing
controversial points of view.

	To this end, local Stanford and Palo Alto news will be
prepared by the staff of the electronic newspaper.



PLAN OF WORK

	The Stanford Electronic Newspaper will be developed on the basis
of two existing systems.  At present the Stanford Artificial Intelligence
Laboratory operates a news service based on the Associated Press A-wire
containing national and international news and will shortly include the
New York Times News Service in the system.  A description of this service
is included as Appendix A.  The service was developed as part of a plan
for developing home computer terminal services along the lines of a
paper reprinted as Appendix B.

	Besides this, the Communication Department with the co-operation
of Stanford Information Processing Center has developed and information
retrieval service for the IBM 360/67 computer called SPIRES which has
been in operation for several years.

	Both of these systems will be used in providing the electronic
newspaper.  At present it looks like it is cheaper to add file storage
to the PDP-10 than to the 360 so the main storage will be there.
The computers will be connected, and a user of either system will be
able get at all the facilities.  The editorial work will probably be
done using the 360, because termials for that machine are available
in the Communication Department, and because they