← Back to search
UNIVERSITY SYSTEM OF MARYLAND
Doing business as UNIVERSITY SYSTEM OF MARYLAND
ADELPHI, MD, 20783
UnclaimedClaim this profile (free)
Prime contracting experience
1 record from government sources. Past participation is not evidence of current teaming interest.
| Role | Agency | Description | Vehicle | Obligated | Action date | Period |
|---|---|---|---|---|---|---|
| Prime | Department of Defense Department of the Navy | MDREN MEMEBERSHIP SUBSCRIPTION | — | $715,886 | Apr 16, 2026 | Jul 1, 2022 – Jul 1, 2026 |
Contract vehicles are shown only when present in the source record. Linked award numbers and vehicles open the authoritative record on USAspending.gov.
Subcontracting experience
1 record from government sources. Past participation is not evidence of current teaming interest.
| Role | Agency | Description | Vehicle | Obligated | Action date | Period |
|---|---|---|---|---|---|---|
| Subcontractor | Department of Defense | WE PROPOSE TO DEVELOP AN END-TO-END (E2E) CROSS-LANGUAGE RETRIEVAL SYSTEM OVER SPEECH AND TEXT, SCRIPTS (SYSTEM FOR CROSS-LANGUAGE INFORMATION PROCESSING, TRANSLATION AND SUMMARIZATION). GIVEN A QUERY, SCRIPTS� CROSS-LANGUAGE INFORMATION RETRIEVAL (CLIR) WILL DETERMINE RELEVANT DOCUMENTS AND SUMMARIZATION WILL GENERATE TEXTUAL SUMMARIES FOR EACH DOCUMENT THAT ENABLE A HUMAN TO DETERMINE WHICH DOCUMENTS ARE RELEVANT.
SPEECH PROCESSING (SP), TEXT PROCESSING (TP), AND MACHINE TRANSLATION (MT) WILL PROCESS THE LOW RESOURCE LANGUAGE (LRL) CORPUS TO PRODUCE TRANSCRIBED, ANNOTATED, AND TRANSLATED DOCUMENTS THAT CLIR AND SUMMARIZATION WILL USE TO DETERMINE RELEVANCE AND GENERATE SUMMARIES. TO ADDRESS THE LRL SCENARIO, OUR NOVEL APPROACH IN SCRIPTS FEATURES IMPLICIT AND EXPLICIT INTEGRATION WITHIN AND ACROSS TASKS LEVERAGING SHARED CROSSLINGUAL REPRESENTATIONS, THUS PRESERVING AMBIGUITY WHICH CAN BE EXPLOITED BY LATER PROCESSES. THIS, ALONG WITH JOINT TRAINING ACROSS CLOSELY COUPLED COMPONENTS, WILL ENABLE THE MITIGATION OF ERROR PROPAGATION ACROSS COMPONENTS. WE WILL USE A FAIL-SOFT APPROACH:EACH COMPONENT WILL IMPLEMENT MULTIPLE, COMPLEMENTARY APPROACHES TO THE TASK, INTEGRATED USING SYSTEM COMBINATION. OUR APPROACHES TO CLIR AND SUMMARIZATION ARE DESIGNED TO OPTIMALLY ADAPT TO A BROAD RANGE OF RESOURCE CONSTRAINTS, ENABLING EACH TO WORK WITH WHATEVER CROSS-LINGUAL RESOURCES ARE AVAILABLE, FROM CROSS-LINGUAL PHRASES TO FULL TRANSLATIONS. OUR APPROACHES ARE COMPUTATIONALLY EFFICIENT AND THUS, WE AIM TO TRANSLATE, TRANSCRIBE AND ANNOTATE THE FULL DOCUMENT COLLECTION WITHIN THE EVALUATION TIME WINDOW, BUT IF THE COLLECTION IS TOO LARGE, OUR PIPELINE ALLOWS FOR PROCESSING TO TAKE PLACE ONLY ON RELEVANT DOCUMENTS.
TASK 1: CLIR. OBJECTIVE: CREATE A ROBUST CLIR CAPABILITY THAT OPTIMIZES THE AQWV MEASURE GIVEN THE AVAILABLE LANGUAGE, SPEECH AND TRANSLATION RESOURCES. APPROACH: COMBINE EVIDENCE FROM MULTIPLE RELEVANCE ESTIMATORS, EACH OF WHICH LEVERAGES A DIFFERENT SUBSET OF THE AVAILABLE LANGUAGE, SPEECH, AND TRANSLATION RESOURCES. UMD PI OARD.
TASK 5: MT. OBJECTIVE: DEVELOP RELIABLE AND FLEXIBLE MACHINE TRANSLATION SYSTEMS WHICH WILL WORK OVER A WIDE RANGE OF LOW-RESOURCE LANGUAGES, DOMAINS, AND GENRES, AND ARE EASILY INTEGRATED IN THE SCRIPTS ARCHITECTURE. APPROACH: WE USE A FLEXIBLE MULTI-PATH APPROACH BASED ON STATISTICAL AND NEURAL ARCHITECTURES. THE SYSTEM WILL USE DISTRIBUTED REPRESENTATIONS TO MITIGATE UNCERTAINTY ACROSS SCRIPTS COMPONENTS, AND WILL LEVERAGE SEVERAL TYPES OF DATA TO COPE WITH LOW-RESOURCE TRAINING DATA, MULTIPLE DOMAINS, AND MULTIPLE GENRES. UMD CO-PI CARPUAT.
TASK 6: INTEGRATION. OBJECTIVE: INTEGRATE INDIVIDUAL COMPONENTS IN A FULL E2E SYSTEM FOR CROSS-LANGUAGE INFORMATION RETRIEVAL AND SUMMARIZATION. APPROACH: WE WILL DEVELOP A MODULAR SYSTEM; WE WILL CARRY OUT JOINT TRAINING OF GROUPS OF COMPONENTS LEADING ULTIMATELY TO END TO END TRAINING OF THE FULL SYSTEM IN LATER PHASES. UMD PI: OARD. 4(GG008712) Parent award context Parent award agency: Department of Defense · Department of the Air Force Parent award NAICS: 541712 Parent award performance state: NY Parent award description: IGF::OT::IGF MACHINE TRANSLATION FOR ENGLISH RETRIEVAL OF INFORMATION IN ANY LANGUAGE View parent award on USAspending | — | $4,686 | Jan 24, 2022 | — – — |
Contract vehicles are shown only when present in the source record. Linked award numbers and vehicles open the authoritative record on USAspending.gov.
Agencies supported
- Department of DefenseGov-sourced
Government-sourced data comes from SAM.gov, SBA and USAspending.gov and is attributed to its source. Items marked company-reported were provided by the company and are verified by FedSub only where labeled. FedSub is a private company and is not affiliated with SBA, GSA/SAM.gov, or any government agency.