|
Anti-Pornography
Filtering
by
David M. Schwartz, CEO, ImaginOn, Inc. 14 October
1998
Filter
Availability
Web content filtering is
a controversial, emotional topic. Many people
consider it an attack on free speech. Others
believe filtering is essential to limit the exposure
of children to "adult" material. ImaginOn takes
no position with regard to the morality or politics
of filtering. Given WebZinger's emphasis on
real-time presentation of graphical data, ImaginOn
makes a state-of-the-art filtering system available
for those people who want one. This content
filter can be selected by checking the appropriate
box on the "Advanced Settings" page of the WebZinger
Recorder Control Panel.
In
the case of WebZinger for Kids, a product for
use primarily by children between the ages of 4 and
11, the ImaginOn Content Filter is always enabled.
This filter is designed to screen out only commercial
sex websites. Websites that are informational, medical
or presenting fine arts are not explicitly blocked.
So, while a search on the word "breast" will not even
start, a search on "mammography" will succeed, displaying
human anatomy and the word "breast", in context.
Understanding that some parents will find this approach
inadequate, ImaginOn provides an additional "open"
filter, under password control. Parents can
add (and later subtract) any words or phrases to (and
from) the open filter.
Filter
Type and Operation
The ImaginOn Content Filter is a proprietary semantic
and spatial filter primarily based on words, the relationships
between those words, and their position within a website.
The filter does not depend on site ratings or lists
of "bad" websites. At one level, the filter
prevents searches from starting, returning the typing
cursor to the beginning of the "FIND" line of the
WebZinger Recorder Screen. At another level,
the source HTML of a website is parsed when the site
is fetched by WebZinger. If a blocked word or
phrase is found on the website, that site is passed
over. Additional "expert rules" beyond words
and word position are also utilized by the filter
to screen out possibly inappropriate sites.
ImaginOn's filter also operates on website pages before
they are displayed by ImaginOn's integrated browser.
A page rejection causes the browser to remain on its
present page.
Filter
Design Goals and Limitations
ImaginOn's content filter is designed to eliminate
99.95% (1,999 out of 2,000) of commercial sex sites
in WebZinger searches where such material is incidental.
WebZinger searches on words or phrases that have obvious
double meanings, such as "foxes" or "girl toys" may
reduce the effectiveness of the filter to 99.5%, (199
out of 200) or less. In practice, this means
that inappropriate images will rarely, if ever, be
displayed in a WebZinger slideshow containing 15 slides.
The
few images that do get through the filter and its
expert rules come from websites whose content does
not conform to the expectations of the filter designers.
Should ongoing testing and user feedback indicate
that this is a problem of increasing magnitude, the
filters will be modified accordingly. Those
revisions will automatically be installed whenever
the user runs the WebZinger Updater Program and a
newer version of the filter is available on ImaginOn's
website.
Inevitably,
some websites with images that should be downloaded
are passed over, due to the bias of the filter.
The rate of such "false positives" is believed to
be less than 10% of the total number of sites visited
by WebZinger during any given search with the filter
turned on. For searches where the search term
is a pun of a sexual nature, the false positive rate
may reach 30%. The filter design purposely errs
on the conservative side.
|