Documents
For Media Mining, the Future is Now!
May 5, 2015
(U) Far Media Mining, the Future Is New!
mam?and
Human Language Technalagy [323)
Run Date:
I) In the first article an the Human Language Technalagy Pragram Management foice's
PMD) activities and plans, we explained that we have five Strategic Thrusts. In this article,
we will facus an the mast active and fast-paced cf the five: Media Mining. Its gaal is ta pravide
seamless access ta infarmatian na matter what the infarmatian's saurce may be -- audia, image, ar
test. Right naw aver twa hundred have access ta same Media Mining cap abilities.
[SllS Near-Real-Tiine Alerts: RT?lil
Integratian af diverse infarmatian saurces ta praduce near-real-time alerts is a majar gaal af
a new Agency-wide pragram, RT means REAL T1ME, and refers ta reducing the time
between callectian and the generatian af actianable intelligence an arder af magnitude in each spin
af the praject.
(SHEET) The first deplaynient af RT?itl ta the in Baghdad in 4th quarter 2006 will facas an
integratian af diverse infarniatian saarces, including GEM vaice intercept and geasgatial
caardinates, ta reduce the time required ta generate actianable intelligence.
New Vaice-S ervices Platfarni: VaiceRT
The HLT PMCI is callabarating with FLT-1D an the develapment af a new vaice services
platfarm, VaiceRT. The first deplayment af VaiceHT, which is architecturally-based an an Army
pratatype knawn as ALICAT, will be aperatianal in the Baghdad nade af RT-lfi in
September 2fifi6. This system is designed ta index and tag 1 millian cuts per day, and pravide
ausiliary HLT services such as language, dialect and speaker identificatian. The cambinatian af
these technalagies with ather RT-lfi capabilities, such as geasp atial caardinates, will pravide a
unique ability ta generate actianable intelligence quickly and accurately.
(SHED Vaice? is a taal that allaws ta perfarni keyward searching an vaice cantent.
{sns Vaice Ward-S earch apabilities
The HLT PMCI's Media Mining Thrust began as an effart ta bring ward-search
capabilities "Gangle far Vaice") ta 1iv?aice Language ta make it easy far them ta
lacate intercept rich in intelligence data. Vaice ward search technalagy allaws ta find and
priaritise intercept based an its intelligence cantent in much the same way as they naw search test
in Far example, in the Glabal War an Terrarism can lacate intercept
dealing with esplasive devices by searching far camman terms such as "aperatian" ar "detanatar,"
as well as mare subtle terms abaut materials ["hydragen peraside"), place names [?Baghdad?), ar
peaple ["Masharaf").
[SllS I) The first generatian af this technalagy has been centered araund Cammercial-aff-the- Shelf
(CUTS) saftware, NEXininer, develaped by a startup campany, Nesidia. The system is designed
(U) Far Media Mining, the Future Is New!
mam?and
Human Language Technalagy [323)
Run Date:
I) In the first article an the Human Language Technalagy Pragram Management foice's
PMD) activities and plans, we explained that we have five Strategic Thrusts. In this article,
we will facus an the mast active and fast-paced cf the five: Media Mining. Its gaal is ta pravide
seamless access ta infarmatian na matter what the infarmatian's saurce may be -- audia, image, ar
test. Right naw aver twa hundred have access ta same Media Mining cap abilities.
[SllS Near-Real-Tiine Alerts: RT?lil
Integratian af diverse infarmatian saurces ta praduce near-real-time alerts is a majar gaal af
a new Agency-wide pragram, RT means REAL T1ME, and refers ta reducing the time
between callectian and the generatian af actianable intelligence an arder af magnitude in each spin
af the praject.
(SHEET) The first deplaynient af RT?itl ta the in Baghdad in 4th quarter 2006 will facas an
integratian af diverse infarniatian saarces, including GEM vaice intercept and geasgatial
caardinates, ta reduce the time required ta generate actianable intelligence.
New Vaice-S ervices Platfarni: VaiceRT
The HLT PMCI is callabarating with FLT-1D an the develapment af a new vaice services
platfarm, VaiceRT. The first deplayment af VaiceHT, which is architecturally-based an an Army
pratatype knawn as ALICAT, will be aperatianal in the Baghdad nade af RT-lfi in
September 2fifi6. This system is designed ta index and tag 1 millian cuts per day, and pravide
ausiliary HLT services such as language, dialect and speaker identificatian. The cambinatian af
these technalagies with ather RT-lfi capabilities, such as geasp atial caardinates, will pravide a
unique ability ta generate actianable intelligence quickly and accurately.
(SHED Vaice? is a taal that allaws ta perfarni keyward searching an vaice cantent.
{sns Vaice Ward-S earch apabilities
The HLT PMCI's Media Mining Thrust began as an effart ta bring ward-search
capabilities "Gangle far Vaice") ta 1iv?aice Language ta make it easy far them ta
lacate intercept rich in intelligence data. Vaice ward search technalagy allaws ta find and
priaritise intercept based an its intelligence cantent in much the same way as they naw search test
in Far example, in the Glabal War an Terrarism can lacate intercept
dealing with esplasive devices by searching far camman terms such as "aperatian" ar "detanatar,"
as well as mare subtle terms abaut materials ["hydragen peraside"), place names [?Baghdad?), ar
peaple ["Masharaf").
[SllS I) The first generatian af this technalagy has been centered araund Cammercial-aff-the- Shelf
(CUTS) saftware, NEXininer, develaped by a startup campany, Nesidia. The system is designed
tb supp art bath real-tinte searches, in which incarning data is autbrnatically searched by a
designated set bf dictibnaries, and retraspectiye searches, in which can repeatedly search
byer bf past traffic. The farmer cap ability the ta functibn as a near real-time
tipp er. The latter capab?ity ta rediscbyer intelligence and tb
refine their search strategies. This can be especially imp brtant in cases where pieces bf a SIGINT
"puaale" became apparent and an analyst needs tb gb back ta preyibus rnessages tb see if ather
unnbticed pieces can be faund.
I) This is very effective because it integrates high-p speech pracessing
with a imp attant agency resburce, analyst bf targets and ntissibns. This
was initially tb the analyst carnrnunity in 2??4 as a
which had been deyelbp ed by Strategy and Gayernance (SSE).
RHINEHART aperates a wide variety bf niissibns and languages, and is used
thraughbut the SAID SS Enterprise. One recent example bf RHINEHART success accurred when
Persian GWCIT searched far the wards "negatiatibns" br "America" in their traffic, and
RHINEHART lbcated a very call that was transcribed yerbatirn prayiding an
an irnp brtant Iranian target's discussibn bf the bf the new Iraqi gayernrnent.
*thes: (U) INSCCIM I US Army Intelligence and Security Carnrnand
(U) Watch far the canclusian bf this lbair: at media mining, casting
tb supp art bath real-tinte searches, in which incarning data is autbrnatically searched by a
designated set bf dictibnaries, and retraspectiye searches, in which can repeatedly search
byer bf past traffic. The farmer cap ability the ta functibn as a near real-time
tipp er. The latter capab?ity ta rediscbyer intelligence and tb
refine their search strategies. This can be especially imp brtant in cases where pieces bf a SIGINT
"puaale" became apparent and an analyst needs tb gb back ta preyibus rnessages tb see if ather
unnbticed pieces can be faund.
I) This is very effective because it integrates high-p speech pracessing
with a imp attant agency resburce, analyst bf targets and ntissibns. This
was initially tb the analyst carnrnunity in 2??4 as a
which had been deyelbp ed by Strategy and Gayernance (SSE).
RHINEHART aperates a wide variety bf niissibns and languages, and is used
thraughbut the SAID SS Enterprise. One recent example bf RHINEHART success accurred when
Persian GWCIT searched far the wards "negatiatibns" br "America" in their traffic, and
RHINEHART lbcated a very call that was transcribed yerbatirn prayiding an
an irnp brtant Iranian target's discussibn bf the bf the new Iraqi gayernrnent.
*thes: (U) INSCCIM I US Army Intelligence and Security Carnrnand
(U) Watch far the canclusian bf this lbair: at media mining, casting