[ analysis of what is heard ]
http://news.nationalgeographic.com/news/2006/02/0222_060222_wiretapping.html
"Rise From the Machines: Surveillance Software Gets Smart"
National Geographic News (02/22/06); Harder, Ben
While lawmakers and advocacy groups debate the legitimacy of the
Bush administration's domestic surveillance program, a host of new
technologies are emerging that could revolutionize the scope and
method of eavesdropping. New data mining programs are appearing
to help intelligence analysts cull through the massive volume of
written, audio, and video communications in search of information
that could be relevant to the war on terror. Advanced applications
can even mine several intercepts at once, offering far more efficiency
and vigilance than human agents, as well as being uninfluenced by
biases and prejudices. Automated sound analysis tools can almost
perfectly distinguish between a child's voice and an adult's, and
can usually determine the speaker's gender, age, and other
characteristics. Determining that the speaker calling in a bomb
threat is a child would call into greater question the seriousness
of the threat than if it were an adult, and programs might one day
be able to determine with certainty the speaker's country of origin,
though SRI International's Venkata Gadde believes that there will
always be limitations to the program's precision. Automated analysis
is also proving more accurate in detecting when a speaker is lying
through the excessive use of 'junk words'--articles, prepositions,
and pronouns, while the use of 'exception words,' such as "not,"
"but," and "except" is usually indicative of honesty, though this
analysis does not come with a guarantee of success, either. Similar
computer programs have been used to gauge truthfulness in news
stories and other writing, and to determine the author of a ransom
note. The University of Arizona's Tom Meservy and his information-system
colleagues are developing a program to analyze non-verbal cues in
video transmissions by analyzing the motions of the speaker's head
and hands, though computers still have difficulty determining whose
body parts belong to whom in a crowded scene.
Received on Sat Mar 02 2024 - 00:57:21 CST