Monica Bozeman graduated from NC State University in 2009 with an undergraduate degree from the Department of Marine, Earth and Atmospheric Sciences. She has worked with the National Weather Service (NWS) since 2012 and currently serves as the NWS AI language translation lead, as well as a project manager for the National Centers for Environmental Prediction.

According to the U.S. Census Bureau, more than 25 million individuals in the country identify as having limited English proficiency. Bozeman said that number could actually be closer to 50 million when accounting for those who may not complete census surveys.

To ensure more people have access to life-saving information from NWS, Bozeman and her team are training AI language models in weather and hazard terminology to quickly translate NWS content into multiple languages. Translated content ranges from hourly weather forecasts to tornado and other severe weather warnings, plus infographics about weather events.

Before Bozeman’s AI work began around 2021, NWS offices in San Juan, Puerto Rico, and Pago Pago, American Samoa, already did translation work, but they relied on manual translations from forecasters. As the roles and workloads of forecasters have changed, however, Bozeman said that they have less time that they can commit to translation efforts. To keep up with how often hurricane information or forecasts are updated, Bozeman said that an entire team of translators would be needed 24/7, which would not currently be feasible for the NWS. 

Bozeman is then leveraging technology to find a solution. Simply using a tool like Google Translate was not an option; Google translation services are globally focused, offering translations that may not reflect the regional differences in a language or dialect that Bozeman’s team wanted to convey. 

Additionally, tools like Google Translate struggled with scientific jargon — something as simple as “rain showers” would be translated to “duchas” in Spanish or “douches” in French, which actually mean “showerheads” or the “act of showering” in a bathroom.

Bozeman sought an AI model that her team could adapt and customize with their own jargon and contexts. AI models that Bozeman worked with early on consistently earned C+ or B- ratings from the NWS San Juan team, but that accuracy skyrocketed when Bozeman switched to using a proprietary AI translation tool from the LILT company. 

LILT’s tool keeps people included in the AI training process. As forecasters translate sentences, the tool predicts the next word that will be used. The tool revises its suggested translation based on the words inputted by the forecaster. Through this training, the AI can now translate NWS content into Spanish with approximately 99% accuracy. The tool can also perform Simplified Chinese, Samoan, Vietnamese and French translations, and Bozeman’s team plans to begin training the model in Arabic this summer.

To cater to regional differences in languages, the AI also lists alternative translations for some words — “lluvias” may be generally understood to mean “rain showers” in Spanish, but in the Caribbean dialect, “aguaceros” or “chubascos” are also common translations. The translation service additionally includes definitions for what constitutes “watches” or “warnings” to better communicate what is at risk for someone who may never have had exposure to such definitions or hazards before.  

In Bozeman’s other role as a project manager, she is leading the software migration for the National Centers for Environmental Prediction. 

All forecasters and hydrologists within NWS use a software called AWIPS to analyze and plot meteorological and hydrological data for weather models. The national centers, however, such as the National Hurricane Center, use a specialized software called N-AWIPS that offers some additional tools. Bozeman is working on software improvements to integrate the N-AWIPS software into one AWIPS platform for everyone to use, which would ease communication between forecasters and emergency managers.

Bozeman said that students can get involved with her work to receive a class credit, or they can get involved through formal internships such as the Lapenta. Her team is particularly seeking students with bilingual proficiency, or with experience in GIS or coding. 

Read about Bozeman’s time in MEAS and her advice for current students in the Q&A below:

How did MEAS help you in your career path?

I knew I never wanted to be a forecaster, so some of those specialized classes that we had at NC State are some of the really fond memories I have that helped shape the direction I wanted to go.

Dr. Aiyyer, his tropical meteorology program was one of the big turning points for me. Beyond just experiencing hurricanes when I was young, actually learning about them and understanding the mechanics turned me in that direction, and I said, ‘That’s what I want to keep doing with my graduate work.’ The Hurricane Center was a thought in my mind of a place I’d want to go, but after that class, I was actually competitive to be able to work at the Hurricane Center because of the knowledge I gained there. 

Another class that sticks out in my mind is Dr. Yuter’s radar meteorology class. That one was a favorite class of mine as well. I feel like everyone should go and take Dr. Yuter’s radar class. All the different colors and the patterns that the precipitation makes day in and day out — to know what you’re looking at, it’s just so important. Radar is available to every single person everywhere, and if you’ve never taken a class to know what it means, I don’t know how you’re interpreting it.

Another one that actually set me on the course of where I went was Dr. Lackmann’s synoptic lab. We had a lab assignment where we had to do programming with shell scripting to make a graphical display for GEMPAK. The National Centers’ AWIPS — we used that in school, and it’s still being used today. I’ve just been working with N-AWIPS and GEMPAK for so long. I had that experience, and so that was pretty influential for jobs and roles that I was assigned through my various jobs.

What is a favorite memory from your time in MEAS?

One is when I was a student trying to do the final presentation in Yuter’s radar class. You were assigned a location, and you had to analyze what would be the best layout of the radars to fit that location. It was interesting to have to be the presenter as the student. 

Then I came back years later. We happened to be visiting town, and Yuter was like, ‘Hey, would you come judge my students’ presentations?’ I went back as a judge, and just interacting with the students and having that experience also influenced me to continue with being a mentor and a facilitator for the NOAA William Lapenta internship. I’ve had an NC State intern from that program.

What advice do you have for students?

We kind of have a changing world going on. The traditional route was always forecasting, whether that’s on TV or as a NWS forecaster, and less attention had been put on the technical skills and the social science skills of a potential candidate for a job. 

But of course, you need to have a lot of different skills under your tool belt. Purely forecasting, purely just the meteorology degree isn’t really going to cut it anymore. If you’re interested in those sorts of things, my advice is to just have multiple skills and multiple hats that you wear.

There’s a lot of good folks who have multidisciplinary skill sets because of little tiny jobs they did or internships they did. So, don’t just look for an internship that’s just meteorology. Expand your horizons a little bit. Do web development, programming, coding, GIS. Social science is blowing up and becoming really huge, especially on the Weather Service side. We don’t have enough social scientists.

Just having more skills under your hat is good to do, beyond the love of weather and forecasting. That way, if you’re searching for jobs, you just have more options, and you’re more marketable because you have these extra skills that you maybe didn’t think you’d use one day. You’ll find out, ‘Oh, this person has skills in this,’ and then you ask him to take a look, and then it becomes a full-blown job. Or, we look to you for this expertise just because you tried something in an internship for three months during the summer. I’ve seen that happen and play out multiple times.

I definitely enjoyed my time at NC State. I’m glad to know that some of the professors that I remember fondly from my time are still working there. They’re all amazing. NC State is a well-known entity within the Weather Service, and when you say that you’ve graduated from NC State, that comes in well-regard, so kudos to the program because they’re turning out some great students.