Broad Operational Language Technology (BOLT) Program
SRI is creating technology to translate foreign languages in all genres, retrieve information from translated material, and enable bilingual communication via speech or text.
U.S. national security requires reliable information that could impact military command and soldiers in the field. Currently, vast amounts of data in multiple languages and formats stream in from around the globe (e.g., conversational speech, email, and text messages). Given the quantity of data, it can be difficult to analyze and determine what’s important. The U.S. also needs to hold conversations with non-English speaking populations of foreign countries.
To help meet these needs, SRI is developing technology for the Defense Advanced Research Projects Agency (DARPA) Broad Operational Language Technology (BOLT) program. Specifically, SRI’s Speech Technology and Research (STAR) Laboratory is leading the research and development of these technologies:
- Accurate, genre-independent translation and information retrieval system
- Machine-mediated bilingual human-human communication system
- Automatic detection of automatic speech recognition (ASR) and machine translation (MT) errors and ambiguities in user input
- System-initiated clarification to elicit user response to repair errors and resolve ambiguities
- ASR for Arabic dialects and two-way speech-to-speech translation between Arabic dialects and English
- Genre-independent translation into English from Arabic (including a variety of dialects such as Egyptian Arabic) and Mandarin Chinese
- Targeted information retrieval from multilingual sources, using natural language English queries with human-machine dialogue for clarification and disambiguation
- Annotation of textual or spoken documents for fast, comprehensive search
- Basic technologies such as parsing, semantic role labeling, language modeling, and other fundamental tools that enable the achievement of the larger research goals
Approved for Public Release, Distribution Unlimited