Current Projects

The Multimedia Language Technologies Group (MLTG) is currently engaged in the following projects:

• IARPA Anonymous Real-Time Speech (ARTS) [2023 - present]:  In support of the IARPA ARTS program, NIST is creating the measurement frameworks to quantify the performance of technologies to anonymize speech, removing identifying characteristics from audio to protect speaker privacy while preserving the actual content of the conversation and the trade-off between privacy (how difficult it is to re-identify the speaker) and utility (how usable the audio remains for analysis). By establishing these benchmarks, we ensure that anonymization tools provide genuine privacy without destroying the intelligence value of the data.
• FBI Voice [2023 - present]:  In partnership with the FBI, NIST is developing the rigorous measurement protocols required to validate advanced voice-based analytical tools. Our work focuses on creating objective benchmarks to quantify the accuracy and reliability of voice recognition systems, particularly in challenging or degraded audio environments. By establishing these standardized evaluation frameworks, we ensure that voice intelligence tools provide a dependable and scientifically sound foundation for federal investigative work.
• Guardians of Forensic Evidence [2025 - present]: The Guardians of Forensic Evidence's initiative focuses on the creation of standardized protocols to ensure the integrity and admissibility of digital forensic data. NIST is developing the measurement tools and validation frameworks required to verify that forensic evidence is preserved and analyzed without compromise. Our goal is to establish a gold standard for forensic reliability, ensuring that the transition from raw evidence to courtroom testimony is backed by transparent, reproducible, and scientifically sound metrics.
• Speaker and Language Recognition (SLR)  [1996 - present]: NIST continues to lead the development of global benchmarks for speaker and language recognition technologies. Our research focuses on creating rigorous evaluation datasets and metrics that quantify the accuracy and robustness of SLR systems to accurately identify who is speaking and what language they are using across different speakers, accents, dialects, languages, and acoustic conditions. By providing a neutral, standardized environment for performance measurement, we enable the development of highly precise voice technologies that are reliable across a wide array of real-world applications.
• Text REtrieval Conference (TREC) [1991 - present]: NIST continues to lead TREC, the premier global forum for evaluating information retrieval systems. By coordinating large-scale evaluation campaigns and providing standardized datasets, we enable researchers and developers to objectively measure the effectiveness of search and text-analysis algorithms. TREC serves as the critical benchmark for the industry, driving the evolution of how the world accesses and retrieves complex information.