Photo By Peter Clute | In partnership with Virginia Polytechnic Institute and State University, Naval...... read more read more
Subscribe
In partnership with Virginia Polytechnic Institute and State University, Naval Undersea Warfare Center Division, Keyport and Naval Sea Logistics Center are exploring how artificial intelligence can streamline and improve the reliability of Navy supply chains. The effort, known as the Sentiment and Topic Analysis for Reliable Supply project, aims to leverage a type of AI known as large language models to improve the accuracy and consistency of contractor performance assessments.
Currently, these assessments include numerical ratings and written narratives that don’t always align, and the discrepancies can make it difficult to accurately assess contractor performance and make informed decisions about future contracts. The STARS project seeks to remedy this through automated analysis of both scores and text, with the goal of producing more objective, reliable evaluations.
The three-year project, which began in 2022, is funded by a grant from the Naval Engineering Education Consortium, managed by NUWC Division, Newport. It is currently in the research and development phase, with researchers at Virginia Tech developing and refining the AI models.
At the heart of the STARS project is a process called sentiment analysis, powered by LLMs. The models are trained on massive amounts of text data in order to learn to recognize patterns and nuances in language that indicate positive, negative or neutral sentiment. When applied to contractor assessments, they use this training to flag instances in which sentiments expressed in text seem to contradict numerical scores.
"The goal is to use AI-powered sentiment analysis to better understand the meaning behind narrative text in contractor assessments," explained Brett Davis, department head of Supplier Quality and Logistics IT Systems at NSLC and the project's programmatic subject matter expert. “If we can improve the quality of narrative text, it provides much more value for applications like supply chain risk management.”
Inconsistencies between ratings and narratives can arise for numerous reasons, including reviewers’ varying writing styles and levels of experience with writing, lack of clarity or specificity in narrative descriptions, and the challenge of conveying tone in writing.
"The AI models don't recognize sarcasm," said Greener. "For instance, if I say [with genuine enthusiasm], ‘This individual was great—they did a great job,' you might infer that I am happy with their performance. But if I use it in a phrase—'Oh yeah, they're great'—and roll my eyes, that wouldn't be necessarily inferred from something that's written down."
Because actual contractor performance data is sensitive and requires special approvals, the Virginia Tech researchers are currently training and refining the AI models using publicly available datasets. These datasets, which include movie reviews and other sources of text with associated sentiment, provide a valuable proxy for contractor assessments while the team works to secure access to the necessary Navy data.
John Greener, NSLC's chief technology officer and product owner for the STARS project, believes it has far-reaching implications. "In the two years since we've been working with Virginia Tech, we've discovered that there may be other applicability for this technology,” said Greener. “This project offers a good opportunity to tackle cutting-edge technologies, with benefits extending beyond the Navy to the DoD and the federal government.”
Greener believes the AI-powered analysis can potentially be applied to a range of acquisition-related tasks beyond improving the accuracy of contractor performance assessments. These could include developing more precise statements of work and more accurately predicting contractor performance.
-KPT-
Naval Undersea Warfare Center Division, Keyport is headquartered in the state of Washington on the Puget Sound, about 10 miles west of Seattle. To provide ready support to Fleet operational forces at all major Navy homeports in the Pacific, NUWC Division, Keyport maintains detachments in San Diego, California and Honolulu, Hawaii, and remote operating sites in Guam; Japan; Hawthorne, Nevada; and Portsmouth, Virginia. At NUWC Division, Keyport, our diverse and highly skilled team of engineers, scientists, technicians, administrative professionals and industrial craftsmen work tirelessly to develop, maintain and sustain undersea warfare superiority for the United States.
| Date Taken: | 04.29.2025 |
| Date Posted: | 04.29.2025 18:42 |
| Story ID: | 496511 |
| Location: | KEYPORT, WASHINGTON, US |
| Web Views: | 48 |
| Downloads: | 0 |
This work, STARS project to boost Navy supply chain efficiency with AI technology, by Frank Kaminski, identified by DVIDS, must comply with the restrictions shown on https://www.dvidshub.net/about/copyright.
No keywords found.
