Title: Voice/Fax User Group Minutes of December 2010 meeting

Release Date: 2018-01-19

Document Date: 2010-12-01

Description: These GCHQ minutes from December 2010 describes NSA advances in using voice analytics – including identifying individual cell towers in Afghanistan where Arabic was being spoken: see the Intercept article Finding Your Voice, 19 January 2018.

Document:

VOICE/FAX USER GROUP
Minutes of December 2010 meeting

b) Google Earth and the B3M VI

The NSA speech research team (R66E) have a framework for the deployment of
voice analytics called HLT (Human Language Technology) Labs:

1 of 4

This information is exempt from disclosure under the Freedom of Information Act 2000 and may be subject to exemption under
other UK information legislation. Refer disclosure requests to GCHQ on , email

< TOP SECRET STRAP 1>



Within this they deploy speech activity detection, speaker and language id algorithms
in much the same fashion as MCS do within B3M. For the forward deployment they
have in Afghanistan, google earth has been integrated into the framework and it has
been discovered that they are able to visualise occurrences where language id (LID)
has identified Arabic being spoken. These occurrences have been mapped using the
geo-location and cell tower information available with voicesail/GSM handsets.
Certain clusters of these identifications have been found to indicate Al-Qaeda training
camps. This unique combination of voice analytics and mapping software has
provided unique and valuable intelligence.

asked the VFUG representatives if this sort of capability would be
useful for GCHQ linguists/analysts if google earth were to be made accessible from
the BVI? While the example above cites language id, it could quite easily apply to
other voice analytics or even non-content metadata should that prove to be of
interest.

c) Speech deployments for December

i. Sorani Kurdish/Pashto/Brahui/Baluchi language id has been deployed against
AQKB selectors for the benefit of the Sorani linguists in OPI~LANG and the Baluchi/
Brahui linguists in OPI~SAAW.

ii. 2 speaker id systems have been deployed. The first is targeting the selectors
associated with a mid-level Baluchi Taleban commander (OBJ BRUMBIES) for

OPI~SAAW. The other has been deployed against the
communications of the Lebanese PM's Chief of Staff for OPI~MEA.

iii. We intend to deploy a further 2 speaker id systems next week. The first targeting

the Director General of Military Intelligence of Rwanda for | in the

OPI~MEA Great Lakes Team. The second against the Lebanese President for^^

also of OPI~MEA.

2 of 4

This information is exempt from disclosure under the Freedom of Information Act 2000 and may be subject to exemption under
other UK information legislation. Refer disclosure requests to GCHQ on , email

< TOP SECRET STRAP 1>

e-Highlighter

Click to send permalink to address bar, or right-click to copy permalink.

Un-highlight all Un-highlight selectionu Highlight selectionh