{ "title": "How Is Human Language Technology (HLT) Progressing?", "type": "SIDTODAY", "description": "This document provides an operational update on the deployment of \"HLT Labs\"—a specialized suite of voice analytics designed to automate the discovery of intelligence \"nuggets\" within massive volumes of audio collection. The document highlights successful field applications in Afghanistan, Georgia, Texas, and Latin America, noting that these tools are becoming \"force multipliers\" that allow linguists to bypass manual listening in favor of automated keyword searching and speaker recognition.", "long_description": "By late 2011, the NSA’s HLT capabilities had moved from research prototypes into critical operational environments. In Afghanistan, the deployment marked a significant technical milestone by utilizing cloud computing architectures for the first time, specifically integrating the DISTILLERY and CLOUDBASE systems on the GHOSTMACHINE platform. These tools supported the full suite of Pashto speech-to-text (STT) analytics, enabling international teams and special operations task forces (such as CJSOTF) to identify target speakers on previously unknown selectors.\n\nThe document candidly addresses the technical hurdles of \"low-resource\" languages like Pashto and Farsi. Because these languages have limited training data, the STT models are less accurate than more \"mature\" languages like Spanish. This is further complicated at sites like NSA Georgia, where noisy VHF radio collections degrade the performance of the algorithms. Despite these limitations, the memorandum notes that speaker recognition remains effective, allowing analysts to geolocate target voices even on unknown frequencies.\n\nIn contrast, the Spanish STT models are described as the most advanced in the NSA arsenal. The memorandum cites high keyword-search accuracy and immediate results in the field: an analyst in Texas identified a drug trafficking lead the morning after training, while a Special Collection Service (SCS) site in Latin America used the tool to rapidly extract intelligence on a Cuban official. The update concludes by noting that these \"HLT Labs\" successes are being integrated into the broader SPIRITFIRE (voice analytics) and TransX (translation/transcription) programs for a full enterprise-wide rollout in 2012.", "tags": [ "media mining", "speech-to-text", "keyword search", "speaker identification", "Special Collection Service", "VHF", "cloud computing", "drug trafficking", "Cuban official", "voice analytics", "linguists" ], "document": { "creator": "NSA", "date_created": "2011-09-06", "classification": [ "SECRET", "SI" ], "dissemination": [ "REL" ], "genre": "newsletter", "programs": [ "HLT", "SPIRITFIRE", "TRANSX", "DISTILLERY", "CLOUDBASE", "GHOSTMACHINE" ], "countries": [ "Afghanistan", "France", "Spain", "Cuba" ], "people": [] }, "publication": [ { "publisher": "The Intercept", "date_published": "2015-05-05", "title": "How the NSA Converts Spoken Words Into Searchable Text", "link": "https://theintercept.com/2015/05/05/nsa-speech-recognition-snowden-searchable-text/", "reporters": [ "Froomkin, Dan" ], "contributors": [ "Fishman, Andrew" ], "doc_source": "https://theintercept.com/document/2015/05/05/human-language-hlt-progressing/", "doc_url": "https://s3.documentcloud.org/documents/2072020/sidtoday-is-hlt-progressing-final.pdf" }, { "publisher": "The Intercept", "date_published": "2018-01-19", "title": "Finding Your Voice: Forget About Siri and Alexa — When It Comes to Voice Identification, the “NSA Reigns Supreme”", "link": "https://theintercept.com/2018/01/19/voice-recognition-technology-nsa/", "reporters": [ "Kofman, Ava" ], "contributors": [], "doc_source": "https://theintercept.com/document/2018/01/19/how-is-human-language-technology-progressing/", "doc_url": "https://s3.documentcloud.org/documents/4352001/2011-09-06-How-Is-Human-Language-Technology-HLT.pdf" } ], "pdf_sha256": "00e9b1a43f1ef11dca4a6eea47f4453898a37f03e05f3695bce7b14a7c969bbd", "pages": 2 }