Usenet filtering service - FREE

[gbmaidsf@wco.com (Ric Sanz)]

From: Charles T. (Dr. Tom) Turley, Ph.D.
      Executive Coordinator - 1WSW
      email:gbmaidsf@wco.com

Dated: 6/27/95
Distribution: World

The following is an exact word-for-word extract content and review of a
service, I was turned onto by Mark Shapiro, Editor of West Coast Online,
my Internet server.

He gave me premission to post it for everybody to review and use, if they
wished. So,I promissed Mark I'd give them a good PLUG, so here goes:  If
you want an easy access, reasonably priced Internet server, then consider
WCO (West Coast Online)  Give em a call at 1-800-926-4683.  There service,
support and expert T1 line connectivity is the best I've ever had.   

This article, which I took advantage of and subscribed to from Stanford
University's Electronic Library project, is FREE, and will save you many
hours of needless wading and reading with your Usenet groups, you may
subscribe to.

For posting and/or FREEWARE distribution with authors premission.
------------------------------------------------------------------------------

Filtered Usenet
(By WCO Staff) Copyright (c)1995 by WCO 


Stanford University's Electronic Library project provides free filtering
of Usenet news for individuals. By sending a simple profile to the
service, you receive only the news articles that match your interests the
service filters out the rest. The news is delivered by Internet email at
no cost. 

How it works

Stanford University builds and maintains a database of recent messages in
Usenet groups. Stanford's database engine compares each message against
its collection of user search profiles. Matching messages are forwarded to
the user. 

Subscribers use the filtering system through Internet email. Anyone can
subscribe by sending a structured message to netnews@db.stanford.edu, or
using the form on the web at http://sift.stanford.edu. Your message is a
series of commands to the automated service. For instance, by sending the
following message you would receive every 5 days excerpts of all Usenet
news messages about online information services. 

To: netnews@db.stanford.edu
subscribe online information services
period 5

The body of each message must include a command to the service. You can
put more than one command in a message, but each must start on a new line
with no leading spaces. If a command is too long for one line, continue on
the next line, starting with a space or a tab. The system ignores the
subject field and the case (upper and lower) of the letters in the message
body.

The service addresses mail to the return address on your request message,
which your email system attaches automatically. Be sure to send requests
from your own account. You can refine your search profile using the
commands TYPE, LINES, PERIOD, EXPIRE, and THRESHOLD. Table 1 lists their
acceptable values. Suppose you request, for the next 200 days, the first
10 lines of articles related to "food recipe" but not "fish" (using a
boolean search). Send this message: 

To: netnews@db.stanford.edu
subscribe food recipe not fish
lines 10
expire 200
end

For the next 200 days you will find recipes in your email. The articles
will be limited to 10 lines, which is enough to give you a flavor. When
you want an entire article, use the GET command.

Big Responses
It saves time to view excerpts of articles. You can tell in a few lines if
an article interests you. This is particularly important when you have to
handle 50 or more messages a day. You can request an entire article
through email using GET, followed by one or more article numbers in the
message body. For example, you want the full text of article 3670 from the
news.announce. conferences Usenet group. Send this message:

To: netnews@db.stanford.edu
get news.announce.conferences.3670
You can retrieve several articles in one message. For example:
To: netnews@db.stanford.edu
get misc.activism.progressive.11965
ca.politics.38420

Best results occur when you learn to tune the search profile to separate
the wheat from the chaff. Stanford's filter engine can help you refine
your profile using message templates. 

When you like an article you read, notify the filter service with the
commands FEEDBACK and
LIKE. The service will use that article as a template for refining your
profile. FEEDBACK identifies the subscription profile and LIKE identifies
the template article. For example, this message says that article
news.announce. conferences.3670 should be a template for subscription 1:
To: netnews@db.stanford.edu
feedback 1 like news.announce.conferences.3670 

The service conducts two types of search: Boolean keyword and weighted
text. You can manually tune your profile with the commands TYPE and
THRESHOLD. Experiment -- If you don't like the results, you can change
settings easily with the UPDATE command.

A boolean search compares the list of keywords in your profile to the
content of articles. Those matching are forwarded to you. For example, the
profile "travel hawaii" in the boolean mode matches articles that contain
both words "travel" and "hawaii." Both words must be present in each
forwarded article but the proximity or frequency of these words is not
considered. You can also use the boolean "not" operator to exclude
keywords.

For example, if you want to read about underwater fishing from any sources
except the newsgroups named alt, use the profile "underwater fishing not
alt."

The weighted text search is more refined. Your profile lists a number of
words and a threshold number. The search engine calculates a score based
on the frequency of these words in the article. If that score is higher
than the threshold number in your profile, you receive that article.

You control the amount of chaff you receive by setting the threshold. Be
careful! Set it too high, and you'll eliminate much of the wheat, too. Set
it too low and your in-tray could be inundated with garbage. Rather than
wait a day, you can test your threshold setting immediately by using the
SEARCH command. 

Searching Yesterday's News
You can search for recent articles already in the Stanford database. The
SEARCH command looks at the database once with the given query. Structure
your profile as you would in a SUBSCRIBE command. For example, to search
for articles related to "information filtering," send:
To: netnews@db.stanford.edu
search information filtering threshold 54 end

Making changes
You can modify your filter profiles with the UPDATE and PROFILE commands.
UPDATE requires the subscription id (SID), assigned to each of your filter
profiles by the service. UPDATE must be followed by one or more of the
commands PERIOD, EXPIRE, THRESHOLD, LINES, TYPE or PROFILE. PROFILE
specifies a new list of keywords. For example, to update the period and
the threshold of your third profile:
To: netnews@db.stanford.edu
update 3 period 1 threshold 45
end

Additional Options
If your email system automatically tags your email signature to the end of
outgoing messages, use the END command to have the news filter ignore any
text that follows. The HELP command returns information on using the
server. CANCEL ends a subscription. LIST shows you all of your
subscriptions.

The Stanford Newsfilter service is on the worldwide web. You can find it
at: http://sift.stanford.edu. There you can fill out filter requests using
a form on the Web page. Some related papers on information filtering
technique are available by anonymous ftp at db.stanford.edu/pub/yan. 

The Stanford library offers a companion filter for computer science
technical reports (elib@cs.stanford.edu), and a search server at
http://elib. stanford.edu. 

Table 1: Commands

Option Value Function
CANCEL SID End a subscription
END Marks end of the command, negates your signature EXPIRE in days
Subscription duration (default 9999) FEEDBACK SID Refine search profile,
use with LIKE GET article ID Retrieve entire article
HELP Call the fire brigade
LIKE article ID Identify template article for search LINES 1 to 60 Excerpt
length (default 20) LIST Lists all subscriptions
PERIOD in days Search frequency (default 1) PROFILE keywords Modifies
profile keywords, follows UPDATE SEARCH profile Do one search of recent
news SUBSCRIBE profile Create a filter request, (use options). THRESHOLD
1-100 Minimum relevancy score (default 60) TYPE weighted boolean Search
technique (default boolean "and") UNSUBSCRIBE Delete all profiles
UPDATE SID Modify profile

Note: The SID (Subscription ID) identifies a particular search profile.
You have one SID for each search profile you create with the SUBSCRIBE
command.



------------verbose signature-----------verbose signature---------- 
Mark Shapiro, West Coast Online, (40,000 monthly copies) magazine and full
service Internet connectivity provider -- info@wco.com, editor@wco.com,
support@wco.com, BBS: 408-946-8592, www.wco.com, ftp.wco.com,
Magazine/Mark S.: v408-946-3823, (Fax)408-946-5331 Internet: v707-586-3060
or 800-926-4683, FaxBack: 707-586-5747 WCO Suite 155 2667 Cropley Avenue
San Jose CA 95132-3707