10B.3 AWIPS Software Integrates AI Translation Technology to Benefit NWS Operations

Wednesday, 31 January 2024: 11:15 AM
336 (The Baltimore Convention Center)
Monica L Bozeman, NWS, Silver Spring, MD; and A. Montanez, J. E. Calkins, K. Farina, and R. Henry-Reeves

The Advanced Weather Interactive Processing System (AWIPS) provides NWS forecasters with the critical visualization and forecasting software platform to support their mission of providing life-saving data, forecasts, warnings, and impact-based decision support services (IDSS). In order to provide the best software tools possible to the NWS forecasters, the AWIPS Program utilizes the latest advancements in technology to meet the Program’s vision: “Creating a flexible architecture that easily adapts to the changing/evolving role and needs of the forecaster.” In addition to providing IDSS and messaging to aid in decision-making during weather events, the role of an NWS forecaster continues to evolve as increased language support services have become a high priority to now be included within the envelope of IDSS. The NWS has a long history of providing product translations, outreach materials and media interviews in Spanish, however many of these efforts have been labor intensive and not nationally supported across the NWS enterprise.

While severe weather impacts everyone, underserved communities, referred to as Historically Underserved Socially Vulnerable Communities (HUSVCs), face a number of barriers, including inadequate communication of weather alerts. NWS is investing in technological capabilities to serve all people regardless of the language they speak and to improve weather readiness of HUSVCs. Therefore, the AWIPS Program has developed the NWS Automated Language Translation Project whose goal is to set the groundwork for developing a nationally-supported, consistent, automated approach to translate English NWS products at scale across multiple languages, Weather Forecast Offices (WFOs) and National Centers. Recent advances in the linguistic capabilities of the Machine Learning and Artificial Intelligence (AI) technology sectors have paved the way for the NWS to begin exploring using these technologies in their operational environment. Since AWIPS is the key platform that facilitates creating and disseminating the weather warnings and alerts, integrating language models trained by AI into the AWIPS software will help the NWS to better serve the approximately 25 million people with Limited English Proficiency (LEP) across the nation. The framework built from the project’s initial translation pilots has proven that high translation accuracy can be maintained while saving bilingual forecasters critical time to focus on forecasting. In addition, this innovation and methodology allows for translation into languages that the NWS does not have existing bilingual expertise nor requires employment of full-time translators.

This talk will provide a brief history lesson on NWS translation capabilities to date, and an overview of the NWS’ current and future methodologies behind the Project’s approach to operational integration of smart translation services. In addition, the presentation will also reveal the benefits the NWS has to gain from using the AI language technology compared to prior translations done manually.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner