(U) Human-Language Technology in Your Future

FROM: Anita H. Kulman
Mission Area Director , HLT PMO (S23)
Run Date: 07/20/2006

[g]
[g

ISERIES:

(U) HLT

 (UI/FOUO) New tools/or analysts to use are on the horizon ...
(U) Would you like to look into the future ? Some visionarie s have been
able to predi ct thing s to come. Throughout the years, science fiction writers
created image s of a technologi cal world which now actually exi sts. A
Robert Heinlein character of the 1940's carried a small , per sonal , portable
telephone , and now cell phone s have become so routine that it's hard to
remember a time that we did not walk around cradling them to our ears.
(U//FOUO) But can you say what the future hold s for the analysts' work
environment ?
(U//FOUO) The Human Language Technology Program Management
Office (HLT PMO ) and the Office of the Senior Language Authority (SLA)
can predi ct that more and better technology services will be available to
more analysts to help them pinpoint the most valuable data within the text ,
voice or image messages that could otherwise bury them. First, let's look
at the HLT PMO and then into the HLT crystal ball to see what the
analysts' desktop might contain tomorrow or in a year or two, or more.
(U//FOUO) The HLT PMO is a relatively new organization that is currently
gathering strength and momentum. It was created to focus research ,
development and deployment of HLT services on one unified path toward
agency mission goals. The PMO , for the most part , fund s other
organizations to do the work of building and refining and delivering HLT
services, but some work is done within its organization as well.
(U//FOUO) The PMO may see into the crystal ball , but its work is not
spectacular magic -- like the smoke and mirror s that you may have
witnessed in some technology demo s, parti cularly by outside companie s
hungry to sell their produ cts to us, where pre senter s promi se miracle s.
The se produ cts often tum out not to work well in our SIGINT environment.
The PMO team works hard to help analy sts do their daily tasks more easily.
It is searching for the sturdy , long-la sting method s that will bring and
maintain solid technology to the entire NSA/CSS Enterpri se.
(S//SI//REL ) In the PMO 's crystal ball today , we can see nine HLT
capabilities that will become more accurate and more useful as well as
more available:
Information Extraction:
IE technologie s organize unstructured data using information from their
content and provide a way to retrieve important element s of those data.
Language Identification:
LID Label s written text or voice messages by language of intere st. (Are
they sp eaking in French or English? l s this document in Arabic?)
Speech-to-Text:
STT provides automatic transcription s of voice intercept in a written form
of the foreign language. (Chinese voice into Chinese text in characters .)

1. Human-Language

2.
3.

4.
5.

6.
7.
8.

IB!

Technology in Your
Future
For Media Mining, the
Future Is Now!
For Media Mining, the
Future Is Now!
.(conclusion).
'Knowledge DiscoverY-':
Finding the Best Material
Human-Languag§
Technology~
EvefY-where
Dealing With a 'Tsunami '
of Interce12t
Building HumanLanguage TechnologYStrangers in a Strang~
Land?

 Machine Translation:
MT pro vides automati c tran slation of foreign language text into Engli sh.
(Chinese text into Eng lish text.)
Optical Character Recognition:
OCR tran sform s an image of a text , in the original language , to a text that
can be edited and searched for pertinent word s or information.
(Transforming a fax of a letter into a Word document.)
Speaker Identification:
SID locates and label s voice messages where a speaker of intere st is
talking. (ls that the terrorist we 'v e been foll owing? l s that Usama bin
Laden ?)
Message Categorization:
Thi s identifie s the type of text or voice message , including the identification
of the topic, of the genre , etc . (ls this a dip lomatic message or an email?).

Information Retrieval / Question Answering:
Information Retrieval -- IR allows analysts to pull important bits of data
from databa ses that contain information of potential intelligen ce value.
Question Answering -- Allows analysts to ask system natural language
questions and to get specific answers, not a list of articles in which the
answers may be found if you look for it. (Where is Islamaba d ? Who is
Musharaf?)

(U//FOUO ) Some of these capabilitie s are already available to analysts,
while other s are still in the first stages of research and may not reach the
desktop for a few year s . The HLT PMO is working on all of them , at
differing levels of resources and funding , and with different predi ction s of
delivery of the capabiitie s to operational offices, within its five major
Strategic Thrusts:
•
•
•
•
•

Media Mining;
Knowledge Discovery;
HLT throughout the Enterpri se;
High Speed-High Volume; and
HLT for Information Sharing.

Look for article s on these Thru sts and other HLT component s, how they
relate to the capabilitie s ju st mentioned , and how they will help you.
(U//FOUO) For more information about these capabilitie s, plea se conta ct
the HLT PMO office (" go HLT" or call
).

[g

II
. '-· ----'·

[g

IComments/Suggestions
I

[i]

11

-

~

I

about this article?

I