DYNAMIC PAGE -- HIGHEST POSSIBLE CLASSIFICATION IS
TOP SECRET // SI / TK // REL TO USA AUS CAN GBR NZL
(U//FOUO) Human-Language Technology -- Everywhere
FROM:
and Anita Kulman
Human Language Technology (S23)
Run Date: 08/22/2006
(U//FOUO) Looking into the future, the Human Language
Technology Program Management Office's (HLT PMO) crystal ball
has given us snapshots of the future for other Strategic Thrusts
such as Media Mining and Knowledge Discovery. But the thrust to
deliver HLT throughout the Enterprise needs a wide, landscape
shot.
(U//FOUO) It is the broadest reaching of the five Strategic Thrusts
because it will provide mature services for all of the nine HLT
capabilities to users not only in SID at NSAW, but to any and all
analysts who need help processing the material that floods their
queues every day. This means that more and better HLT will be
deployed to analysts at the Cryptologic Centers and field
sites in the US and throughout the world.
(S//SI//REL) For some HLT services throughout the Enterprise, the
future is now: many analysts are already using HLT and may not
even be aware of it. At its best, HLT functions "upstream" as close
as possible to collection sites, and definitely "behind the desktop."
When this occurs, analysts receive queues of messages that have
already been sorted and filtered by some HLT service that may
have labeled the messages with information to alert the analysts to
possible good intelligence or to tell them that the material is
probably junk.
(S//SI//REL) When HLT is visible to analysts, they can see the
services at their disposal at a touch of their fingertips by accessing
a website or by hitting an option on a tool already on their
desktops. There are speech, text and images services available, as
follows:
Speech Activity Detection (" Is there speech in this cut? ");
Language ID (" Is this conversation in a language I'm
interested in because I've found intelligence thre before? ");
Speaker Search (" Is this the speaker who has provided
good information in the past? ");
Machine Translation in over 70 languages (" Does this text
have information that should be translated carefully by an
language analyst? "), and
Optical Character Recognition (OCR) for most Roman,
Cyrillic and Arabic scripts, which turns images of text in
faxes and other electronic messages into searchable text.
(S//SI//REL) These tools are available right now, in the present,
and the near future promises a lot more activity. For example, for
OCR, new languages are being enabled monthly and training on
how to use OCR tools will be forthcoming. The PMO has invested in
an "OCR-on-demand" capability. In addition, keyword search of
SERIES:
(U) HLT
1. Human-Language
Technology in Your
Future
2. For Media Mining,
the Future Is Now!
3. For Media Mining,
the Future Is Now!
(conclusion)
4. 'Knowledge
Discovery': Finding
the Best Material
5. Human-Language
Technology -Everywhere
6. Dealing With a
'Tsunami' of
Intercept
7. Building HumanLanguage
Technology
8. Strangers in a
Strange Land?
OCR printed fax is available for Chinese and is being studied for
most Roman, Cyrillic and Arabic scripts.
(S//SI//REL) As another example, currently available speech
services, speech detection and language ID are being integrated
into BABBLEQUEST and are being made more robust. This means
that they will need less maintenance and will deliver good
information consistently over time. Comparison and Speaker
Watchlist services are being added to the Agency operational voice
architecture as services available as on-demand pull-down menus
from HOTZONE .
(U) Watch out for these innovations!
"(U//FOUO) SIDtoday articles may not be republished or reposted outside NSANet
without the consent of S0121 (DL sid_comms)."
DYNAMIC PAGE -- HIGHEST POSSIBLE CLASSIFICATION IS
TOP SECRET // SI / TK // REL TO USA AUS CAN GBR NZL
DERIVED FROM: NSA/CSSM 1-52, DATED 08 JAN 2007 DECLASSIFY ON: 20320108