How Better Translations Could Save Lives During Disasters

In this episode of The Alerting Authority, Jeannette Sutton and Eddie Bertola sit down with Dr. Joseph Trujillo-Falcón, a leading researcher in multilingual risk communication and emergency alerts.

Joseph shares his journey from immigrating to the United States from Peru to becoming one of the nation's foremost experts on language equity in weather warnings. The discussion explores how emergency messages are translated, why certain warning terms can create confusion, and how artificial intelligence is helping agencies like NOAA and the National Weather Service improve multilingual communication.

The conversation covers:
• The challenges of translating tornado warnings and weather alerts
• Why direct translations often fail during emergencies
• AI-powered weather warning translation systems
• Language equity in public safety communication
• Wireless Emergency Alerts (WEA) and multilingual messaging
• How local agencies can better reach diverse communities
• The future of emergency communication for the 2028 Olympics and beyond

Whether you're an emergency manager, public safety professional, meteorologist, researcher, or simply interested in how technology can help save lives, this episode offers valuable insights into the future of alerting and warning systems.

Guest: Dr. Joseph Trujillo-Falcón

Topics: AI Translation, NOAA, National Weather Service, Emergency Alerts, Risk Communication, Public Safety, Weather Warnings, Language Equity, Wireless Emergency Alerts, Emergency Management.

This episode is sponsored by HQE Systems. 

#EmergencyManagement #PublicSafety #RiskCommunication #AITranslation #LanguageAccess

Transcript

SPEAKER_00 0:06

Hi, I'm Jeanette Sutton.

SPEAKER_07 0:08

And I'm Eddie Bertola.

SPEAKER_00 0:10

And we welcome you to another episode of the Alerting Authority. And as always, we encourage you to subscribe, follow, listen, and most importantly, participate in these podcasts. We want your questions, your concerns, ideas, problems, pain points, and your success stories so we as alerting authorities can better do our jobs and make our communities safer.

SPEAKER_07 0:32

And we appreciate all the sponsors. And if you are interested in being a sponsor, please reach out. This episode is sponsored by HQE Systems, and they are an IPAUS-approved alert origination software provider, which means it can support communications and organizations that need to send emergency alerts through authorized public warning channels. One of the strongest parts of their approach is that they're not just offering technology, they're helping organizations bring several systems together to make it easier to manage so that when you have those disastrous situations that we all have, you can organize everything and manage all your communication in a single piece of software. So just big thank you to HQE Systems.

SPEAKER_00 1:15

Today we have Dr. Joseph Trujillo Falcone, who is an assistant professor at the University of Illinois Urbana-Champaign in the Department of Climate, Meteorology, and Atmospheric Sciences with an affiliation in the Department of Communication. He leads the Advancing Language Equity and Resilience in the Atmospheric Sciences, or Alertus Lab. I'll say that again, Advancing Language Equity and Resilience in the Atmospheric Sciences, because you're going to want to look that up again later. He leads that lab, which works to improve how extreme weather information reaches the public in multiple languages through community and practitioner engagement. His interdisciplinary research has driven policy changes across the National Oceanic and Atmospheric Administration and the National Weather Service. Closer to home, Joseph partners with emergency managers across Illinois to advance public safety, work recognized with the 2026 Community Impact Award from the Illinois Emergency Services Management Association. Together, these contributions have established him as a national leader in weather risk communication. He is also an award-winning bilingual multimedia meteorologist. He has worked in radio and broadcast networks and continues to serve the public through My Radar, a weather application used by more than 50 million people worldwide. He also serves as a science contributor to the Illinois Public Media PBS and NPR podcast, Weather Realness. He holds a BS in Meteorology, a BA in Spanish from Texas AM University, and he earned his MA and his PhD from the University of Oklahoma. I think we met when he was a graduate student. Oh my gosh, that's right. We have to go back farther. He wasn't. You didn't even list.

SPEAKER_07 3:12

I mean Illinois, right? But that's not your alma mater.

SPEAKER_00 3:16

Pre-Illinois, pre-Oklahoma, going all the way back to Texas. And I was at Kentucky, and I tried so hard to get Joseph to leave Texas and come to Kentucky because, as you will hear today, he is just a phenomenal scholar. And his research space bridges atmospheric science, meteorology with expertise in communication, and he is bilingual and brings all of this together in this just amazing research agenda. So I am so happy to have you here today, Joseph. Welcome to the Alerting Authority.

SPEAKER_04 4:01

Thank you so much. As a fan of the podcast already, it's such a pleasure to be here.

SPEAKER_07 4:06

Well, I have to say, Jeanette, I you didn't get him to accept his way like into the transfer portal for academics between universities. That that has to become a thing now. I mean, if it can be done for athletics, to to pick and grab amazing research and amazing scholars and bring them to your university, I'm just saying let's let's start a whole nother little uh thing for the university system.

SPEAKER_00 4:31

Oh, we do. We try. We try to poach students because we've got projects for them to work on. I think that the remaining in Texas was the better option for Joseph at the time, but then he made his way to Oklahoma and again chose not to come and join me. But I got to be on his dissertation committee, so I still got to participate a little bit in his amazing research that I'm positive he's gonna tell us about today. So um, yeah. So Joseph, so I shared a little bit about you, but I think we need you to tell our audience. Um, you know, share with us who you are and how did you come to study alerts and warnings in this very special area that you focus on, and you're one of very few in the United States, maybe in the world. We're just so privileged to hear your story today.

SPEAKER_04 5:27

Now thank you, Jeanette. Thank you, Eddie. So excited to be here today. And really, the origin for this all just started out in my own upbringings. I was born in Lima Peru and immigrated to the United States when I was five years old. And so naturally, I had to adapt really quickly to a new culture. Anywhere from the food that we ate, you know, I didn't even know what uh uh, you know, a hamburger fully was until I came here to the country. And then, of course, the weather disasters that came my way too. I grew up in Dallas Fort Worth. And, you know, Lima Peru, I would characterize as a climate very similar to like California, you know, coastal. We're usually calm for most of the time. We'll get the occasional sprinkle too. Uh, but you know, I came to Dallas Fort Worth, and to me, the sky was exploding. You know, I've never experienced a thunderstorm like that ever, and hail and tornadoes. And suddenly I was the one translating to my own immigrant community growing up uh how to stay safe during these kinds of hazards. And uh I was actually personally scared of storms. And it actually there was one time where my mom had to grab me by the ears and drag me outside during a hail storm to be like, you're not gonna be scared about this anymore. And that's when I realized how beautiful Mother Nature really was, and that's what got me interested in meteorology. And my initial goals was to go out and be a bilingual broadcast meteorologist, to give back to the communities that made me the person I am today. But time and time again, I kept experiencing language inequities. I remember vividly, you know, uh uh doing an internship for Tele Mundo and asking them what a certain word meant in Spanish, because I was learning everything in English, so I didn't really know how to say certain scientific jargon and terminology. And they were like, well, that actually that word just doesn't exist, so just say it in English. And I was so baffled by that. Like I think of my mom, I think of my community, and they wouldn't be able to understand that. And that frustration led me to 180 into the work that I do today, identifying language inequities and really being able to use community insights to advance communication and tactics that work for everyone. Because I'm a fond believer that everyone deserves to receive life-saving information at the end of the day. Yeah, I love that.

SPEAKER_07 7:35

Now, are you gonna give your mom a shout out? Oh, 100%.

SPEAKER_04 7:40

Her name is Amorista Falcon, and she 100% uh inspired me to get into the field of meteorology. Uh, and even to this day, you know, it's uh, you know, she's still every time I do a survey or design a study, I always run it through her. She's like the lead PI of all PIs out there. So huge shout-out to her.

SPEAKER_00 8:00

Thank goodness for family members who have their little fingers in our research because they're like the extra layer of like uh verifying that what you're asking isn't crazy, and they give you that feedback of I don't get what you're asking, and you know, fix it.

SPEAKER_07 8:17

For sure. So, and props to her for taking you outside. Um, and you think about all the little actions that we go through in life. I wonder if she knows how many people she has impacted through you, through those types of decisions, right? By just helping you to face a fear and then not only facing it, but then saying, I want to become the master of this fear and and just doing what you do. So let me just throw out the next phase here because we're gonna get into research, because you didn't just stand in that hailstorm and say, Wow, you know, pretty. You know, it it really got your mind going and it has led to a bunch of different research. And as you talked about, language inequities, uh, that is something that we experienced and still experience in California all the time. And if you've heard our show before, which I know you have, we really believe that everybody deserves to get information in the language and manner that they can consume it in, right? So that they can take that action. So if we jump into some of your research for tornado warnings, you've looked at different variations in Spanish of types of warnings that go out. Can you just I mean, and tornadoes are one hazard, but it's one that now I live in Arkansas that's unfortunately now very near and dear to my heart. What have you learned and how can it be applied into practice?

SPEAKER_04 9:42

Yeah, no, you know, this study that really looked at tornado warning terminology really did stem from a lot of that experience working with communities firsthand. Because, you know, uh I grew up learning the difference between watch and warning, and even to this day I have to remind myself because it is full of jar sometimes. And, you know, warning being the more imminent, go take shelter in the watch, being like, let's look out for, you know, uh possible conditions that may develop over the next couple of hours, for example. And the thing is, is that in Spanish, there's not really a way to directly translate that. And so when you lose that context and that meaning, you may not even be prompted to consider it as dangerous. And I experienced this firsthand when uh, you know, the official translation of the word tornado warning from the National Weather Service, even to this day, is aviso de tornado. Well, in Spanish, the word aviso is really more of like, I kind of advise you, kind of like a pat on the back. Like, you know, me, my mom would me avisaría to go brush my teeth, uh, rather than like, you know, like, hey, this is a recommendation, suggestion kind of thing. Not like you, you need to right now. And, you know, that was felt across the entire bilingual community. Bilingual broadcast meteorologists were saying all the time, this isn't we keep saying this, but it's not as urgent. And actually, uh, this goes all the way back uh even to some research done in the late 80s. Uh, Dr. Aguirre, who used to be at the University of Delaware, actually conducted a study in Saragoza, Texas, where uh tornado was approaching uh the their community, and more than half of the people that died that day were Spanish speakers. And it was because a radio announcer at the time was trying to communicate that in Spanish to the community. Uh, but uh they he didn't have a way to translate the word warning, and so he tried to make something up on the spot, caused a lot of confusion and language inequities. That was the first time in a government report that language inequities were linked to uh fatalities and casualties. And so we wanted to address that once and for all with a nationwide survey of Spanish speakers and test the variations of the term tornado warning, because again, there's no direct translation of that word, there's various ways of saying it, maybe alerta, maybe advertencia, emergencia. And we just wanted to ask people how they perceived the risk among these different kinds of words. Well, what we ended up finding is that um aviso ranked among the lowest uh when compared to all the other variations. And uh even the word advisory in Spanish, advertencia, was ranked as more urgent than warning in Spanish. And we, you know, some eye-popping data here to show that sure, we were translating the word, but not necessarily the meaning of that word. And really tried to have an open conversation about when we try to expand to other languages, it's not just an easy thing of going into Google Translate and calling it a day, but understanding the contextual meaning of language, the dialects, the culture, the contextual and rich background that is necessary to prompt people to take into protective action. And all of this, of course, also goes back to what uh Jeanette has really been saying over the last couple of episodes, too, about plain language communication, that we wouldn't even have a lot of these issues to start off with if it was plain language in the first place. But we were able to uncover a lot of the nuances that currently exist in our communication system.

SPEAKER_00 12:58

Do you ever wonder like who was who was sitting at that desk when they made the decision that this is the word we're gonna use? And like being able to go back, you know, a couple of decades to help them to figure out how to make that word much more directive may have saved some lives. You know, if uh Aguirre's research really pointed to that, then like boy, just a word here and there could make a huge difference.

SPEAKER_04 13:25

No, for sure. And you know, uh, you know, hat tip to Aguita there for uncovering some of that. And that that was the thing that I, you know, frustrated me because like the research found this decades ago, but we haven't been able to put it into practice quite yet. And they're really, you know, within gov uh government agencies for multilingual communications, really more grassroots. There's been a lot of admirable and amazing uh volunteers across the National Weather Service that translate um, you know, on their time, no additional pay or anything. They just do it out of the kindness of their hearts. Um and while we, you know, definitely admire that, it's all been grassroots. So there's been no way to actually organize and set established ways to go about doing this. Like even to this day, there's no official weather and climate dictionary in the Spanish language uh across the country. So if we can't even describe that word to somebody, how are we gonna ever expect them to take it seriously and you know eventually take protective action?

SPEAKER_00 14:20

So Yeah. Yeah. I don't want to jump too far ahead in our interview, but I wonder if that's gonna be one of your future projects is to create a dictionary like that. Because I can just I can just imagine how valuable that would be to people in the in the field, in the broadcast industry, in the research space. Um yeah. Um I okay, so let me move on to be you as we were starting to talk a little bit about wireless emergency alerts, and I want to move there because we know that um wireless emergency alerts, they are those messages are pre-created by the National Weather Service. So they're already pre-written. So um I want to know have they included the word alert us in those messages? But I also want to hear about your dissertation research and the importance of WIAs that are delivered to limited English proficiency populations and what you learned there, because that was a fantastic study.

SPEAKER_04 15:20

Thank you so much. I, you know, in terms of the word alerta or aviso, we put that study out there and, you know, as a lot of things um that have to get processed through through the government, it's currently being evaluated. But we have been able to pass previous terminology uh changes in the past, like the Storm Prediction Center's risk category outlook. That kind of gives you a heads up of what uh severe weather threat is incoming in your direction. We've been able to change it in Spanish before we ever got to the English version. So that was a huge victory on our end of really the first research to operations effort for the ADC that was done completely in Spanish. So we're looking forward to continuing that conversation for the alertas. But in the meantime, uh you know, the National Weather Service back in 2020 actually uh started issuing these Spanish language Wiias alongside the English version. And they've through those volunteer translation teams I was talking about, they were able to automate them to where, yeah, if you received um on your phone a English language WIA and one, and you also had Spanish set as either a primary or secondary language on your phone, you'd get a Spanish alert uh WIA immediately following the English one. And that was done across the country and it still happens to this day for at least weather WIAs. Um beyond that, I cannot speak on half of that. But weather Wii as we've been able to see some of those advances, both in the 360 version, the 90 character version. And so, you know, we just really wanted to get an understanding because again, oh, admirable work for the for from the weather service to go out there and automate this, but we really wanted to see how people resonated with the message and what better ways there are out there to improve some of this work. And so we um interviewed uh about 27 Spanish speakers from across the country. Uh, and we actually got them to read the Tornado WIA out loud and verbalize their thoughts in real time. And we applied, you know, these foundational warning communication uh models like the warning response model and protection motivation theory to see how people received understanding and really made sense of these WIAs in real time. And, you know, a lot of findings already came out to the bat. Well, I mean, first off, people again saw the word Aviso and were like, this is not urgent enough for me. So just, you know, kind of validating the previous work that we have done. But we were able to go much deeper into the contents of a WIA. Some that really is stuck out to me were acronyms. You know, we there's so many acronyms in the warning uh communication landscape and so many agencies that it could even be confusing in English, but you try to translate that into Spanish and it uh it's already all over the place. So uh in the 90-character version, uh the word National Weather Service gets uh uh reduced to an acronym, NWS. And so the Spanish version translated that to the word Servicio Nacional de Meteorología or SNM. And so in the 90-character version, the source is usually the first thing we see, and so it said SNM. And as people were thinking aloud this message, I asked the people were like, What did what does that even mean? Um, and so we got some really funny responses. We got uh someone thought it was a seismology service. I was like, okay, someone thought it was an MM, like uh the candy. Uh, someone else thought it was an SMS message and not even relating to an agency whatsoever. And then my favorite one was uh someone thought it was one of Rihanna's songs, SM, which I won't go deeper into. And related it to media some. And I was just like, okay, well, we have a lot of varied responses here. And uh, oh yeah, and also neuromolective, uh neuroleptic malignant syndrome was another one that I thought was interesting. And so at the very beginning of the message, people were already pretty confused in the 90-character version. And then we, you know, uh each one of these also have time zone abbreviations like CST, EST. And so um, I remember uh for PSD Pacific Standard Time, someone was like, What they were thinking out loud, like, does this mean priority storm time? And I was like, interesting, interesting. So we're just out here taking notes. And then of course the dialectical issue also kicked in. There were other examples where like the word debris doesn't directly translate just like the word warning it does in Spanish. And so there are different ways to say it. Um, and one of the ways that the Weather Service initially translated it to was the word proyectiles to kind of indicate some flying debris up in the air. But proyectiles can have various meanings depending on where participants grew up, especially in parts of Central America where there's has been history of political strife in the area, proyectiles doesn't really mean much more debris than actual flying projectiles in the air. And so some people were equating that word to there's a tornado and missiles flying in the sky at the same time. And so we got to learn a lot of how dialects really kicked in there as well. And I guess most, you know, really digging into the final parts of this study too, uh, recently arrived immigrants actually really used Wii as overall. They thought they it was useful, you know, for all these uh sorts of feedback they were giving. They did prefer the longer 360-character version because it produced that necessary context for them to understand this because we have to take into account a lot of people that recently moved to the country may not have experienced tornadoes before. And so even if it's a small additional amount of characters, they were able to contextualize it, know what protective actions to take, and feel more motivated to eventually to take that protective action. And so uh yeah, this piece is almost out. Uh, we just submitted the last revision there to the Journal of Applied Communication Research and hopefully get to open up a conversation because when we look at the warning response model and look at those five warning components, technically the Wii has met nearly all of them. But going beyond that and looking at the cultural and linguistic understanding of the message to really get us one step closer to engaging multilingual groups in the future.

SPEAKER_00 21:08

It's so important. And your research is truly gonna open a lot of doors for more. And I know you have graduate students that are now working with you to help to push this space forward. Um Yeah, when we when we get into later in our in our conversation, when we talk about AI, this contextual issue that cannot be translated by a system is going to be a real issue that I'd I want to dig into with you.

SPEAKER_07 21:38

For sure. So and I'll say too, uh the National Weather Service has made a lot of progress with the translation into Spanish. And that is something that is supported through, you know, FEMA iPAUS, right? The for the wireless emergency alert. Um but that is currently the only language that is supported outside of English. And And so it is a huge accomplishment. But we're not just English and Spanish speakers, right? We have so many different languages. But what I'm hoping is the groundwork that you're laying through your research and through the recommendations that are going to come from this can then be applied to other languages. And again, it there will always be nuances, but I mean this this is the hard part. And you are you have been working really hard with it. And I can say, as someone who has sent a lot of messages, we don't have translation services available right all the time. And I mean, you can look at some large agencies, they they may, but I I look at it from a national landscape, right? I mean, there there are small entities out there that are doing the very best they can and they're they're putting it out in English. And A, I mean, I'm I'm just happy if it's not confusing in English. Um, because as you've seen, examples, sometimes it's as an English speaker, I don't even know what they're talking about. And then I'm thinking to myself, okay, give them a shot to translate it into Spanish. Um, so I think for my for my question is we're we're starting to look at at AI. AI is something that that exists, and AI is this really big umbrella. I know we're saying AI in a general term. We'll get comments. There are so many things about artificial intelligence, but and how it's applied to alerts and warnings. Now, I know you've been working with NOAA, National Weather Service, to look and see how this can be done. So, with all of your experience here, what's your opinion on the use of artificial intelligence to help create and just in the space of alert and warnings?

SPEAKER_04 23:50

Yeah. Uh, you know, as we start really, you know, going into this, or where we're already in in this era of artificial intelligence, I do think it can be a great tool to assist you in translating if done correctly, and if it has quite a amount of human input along the way. One of the very first studies that I did before even working with the National Weather Service was just run the, you know, general Chat GPT models and gave it an English language forecast and said, you know what, let's put it on here and let's see what it spits out. We uh uh published this in the journal Artificial Intelligence for the Earth Systems. And surely enough, a lot of the mistakes that I was noting in the WIA study were starting to pop up here. The lack of uh there was translating the official term for uh warning of Abiso into different other variations that the weather service wasn't approving, thus increasing the inconsistency if multiple people do it at the same time. It wasn't really breaking down the hazard in a way that was consumable to the public. And if it didn't really have a word, it would really either make one up or try to explain it in a different way that didn't really make sense. And so, you know, at first hand, you know, it a lot of people initially thought that with the emergence of LLMs, we could just easily put uh uh something in English and spit it out into multiple languages all at once. But here's the thing uh, you know, machine learning in general has been actually pretty much developing a lot in the language world for years now. I mean, when you think about Google Translate, that's really what it really is driven by a lot of those models inherently. And the thing is, is that uh in order for these uh kinds of tools to be successful and for them to be able to be applied in such a context, really the machine learning in general needs to be more acquainted and adapted to that context context, specific context in hand. So that's what we essentially strive to do with NOAA. Like if we put uh a weather warning in Google Transit right now, it wouldn't be able to do it right because it doesn't understand the true context of weather and climate, and then of course doesn't understand the terminologies and things that they have to consider. Um and so for that reason, uh we uh worked with a private sector company called Lilt that had a neural machine translation framework. And essentially, what it we gave it the entire weather service domain to give it all of the terminology, all the things that the weather service uses on a daily basis. And the machine then produces a first guest translation of it. But whenever that first guess translation came in, we spent over three years training the model with bilingual forecasters from the National Weather Service, certified translators from the American Translators Association, and other scientists too, to make sure it was grounded and it understood that proper context. Because there were some really funny instances I have to share with y'all. Um some of the more simple ones is it would confuse like tornado watch for the watch you have here on your wrist. Um in Chinese, uh, the word um atmospheric trough translated to a trough where pigs eat. And like um one of my other uh favorite examples to, I think it was in oh my goodness. Oh, well, yeah, this is another Spanish example was the word rip currents. Depending on where you're from or how you're raised, you can say it differently. It initially said the word resaca, which actually means hangover for a lot of Latin America. So it was issuing hangover warnings for the beaches, and we're like, ooh, probably don't want to do that.

SPEAKER_07 27:15

But still applicable in some beaches, so it can, right?

SPEAKER_04 27:19

Right, right. Spring break. And so yeah, no, for sure. And but we were trained over time to make sure it would really understand that context. And we were able to achieve up to 99% accuracy for some weather service products and reduce translation time for the weather service from one hour to two to three minutes for some products as well. To where it, you know, we really move forward now in um in making sure that, sure, it creates a first guest translation using all that contextual knowledge, but at the end of the day, there's still a human forecast so that can look at it, verify it, and send it out. We were inspired a lot from models like Environment Canada, uh, where they act, you know, it's it's law, federal law whether that they have to issue stuff in English and in French. And so how they do it is they make an English language warning that comes in templates. That's a discussion point, um, and then get sent over through automation to a translation office that also works 24 hours a day, seven days a week. And they, you know, in a matter of 10 to 15 seconds, English followed by French comes out. And so, you know, hopefully this is the first step of many to reimagine how that could look like here in the United States, building up that automation that could help us uh mend that bridge and maybe one day have that translation office here too to spread it to other languages. But so far, so good though. We've been able to reach uh this technology has the capacity to reach over 30 million limited English proficient speakers present day. And it's now even being used here in the World Cup, as we have uh, you know, over seven FIFA official languages that they use. And so we've used that as an opportunity to expand some resources for them too.

SPEAKER_07 28:56

See, that's exciting. And and again, it's one of those if you work in a small agency and you need to translate something into Spanish, um, hey, it's it's not required, right? If you're gonna send out a wireless emergency alert. But knowing your community, I mean, if if you have a percent of your community that's not gonna receive it and and understand it, well then what are you gonna do? Right? You you need to think through this. And that's where again your research and these practices come into play. Because I I do know of some that will just put it into chat because they're like, it's better than nothing. I don't, I don't know what else to do. So I'm either gonna give them something that's okay or just not send them anything at all. And that's that's where a lot of people live right now, and it's uncomfortable. It is not the what we want. But that's where your experience and your research, I think, is helping to develop I think the solutions that are gonna be the long-term like path forward.

SPEAKER_04 29:55

For sure. I think this opens up conversations for how we should be approaching this, uh, the methodologies for how to properly train these systems and also evaluate them too, because you know, we have to constantly be training this model or else it's gonna forget and it's gonna hallucinate over time. There's actually one example, thank goodness it didn't ever get released, but in a hurricane uh key message or one of the public uh oh my god, the tropical weather outlooks, it was it accidentally hallucinated to saying like chewing pink bubblegum in the middle of a of that. And we're like, where did that come from? And so we actually have metrics set in play for this AI translation program where I've worked a lot with sociolinguists and uh language studies uh scholars that they have ways to actually grab uh a pair of words to evaluate how the meaning gets translated and what exactly gets translated. There's metrics that actually evaluate that, and there's various different kinds. And so we ran, we actually have an automation that runs those metrics on all the translations that are being developed. And if it ever reaches a threshold that's not good, it pings to us saying that we need to retrain the model more in tornadoes before we something needs to be released. And so we're still experimenting with it. This is still brand new, but we're really trying to find that sweet spot, that Goldilocks out, of course, of uh making sure the language is accessible and effective, but also making sure that it's of high quality. Because we uh, you know, this different languages have different kinds of rules to and syntax to them. And we uh, you know, don't want to just disrespect people's languages and and practices because of that too. And so yeah, it's been a really interesting way to kind of explore how we we can apply AI in an ethical and effective way to reach more people across the country.

SPEAKER_01 31:42

I think that's amazing.

SPEAKER_00 31:44

And um, I want to know, can you tell people how to actually access that information if they want to see it? Um, the app that's now available.

SPEAKER_04 31:52

Yeah, of course. People can go to weather.gov slash translate right now and be able to access a lot of that information. Uh there are uh so you know, there's 122 uh jurisdictions of the National Weather Service across the country, and we've been slowly expanding to all of the jurisdictions. I think we have over 40 of those jurisdictions covered, anywhere from Atlanta, Georgia, all the way out to Portland, Oregon, uh all the way even to Seattle. Um, we even have like um, you know, national centers like the Storm Prediction Center, the National Hurricane Center products there too. And so for those regions, they can access in real time. Like if, say, one of those offices were to issue a tornado warning on the spot right now, um, you would get it in real time, uh, the translation right away. And in the textual products, we actually have something, uh, everything's just usually in black text, but there is stuff in blue text that you can click on and it gives you the dialectical variation of the word. Uh, something that we took into heavy consideration here was we didn't want to endorse one dialect over another. That would, you know, be disrespectful to communities, and we don't want to say that this is the correct way of speaking Spanish, for example. And so we leave it up to the user that, you know, if you're an emergency manager in Florida that serves a lot of Cuban and Puerto Rican populations versus someone in Dallas that's serving more of a Mexican population, here's the alternate ways of saying it. Um, but there's probably there's also you can uh have access to visual products. So all of the warnings right now, like all the heat advisors as of time of this recording that are all across the country, the visual aspects are also translated into Spanish and up to 12 other languages as well. Um, and then most importantly, that's something that anyone can use today at weather.gov slash translate are infographics that we have for all the various kinds of weather hazards from tornadoes to heat to air quality to wildfire. And we have the English version there set up for you, but where when you click on the picture, it gives you that a version in English and then social media captions that meet the Facebook and and uh Twitter slash X character counts. And if you access it in Spanish, the same thing will pop up to where you don't even need to speak that language. You can just copy and paste it right before a a big extreme weather event, and you can let your communities know of what protective action they should be considering. And so we, you know, we we've been trying to deliver the full model here, not just the communication aspect, but the education aspect as well, especially since these multilingual groups don't experience uh or may not have experienced those kind of hazards in their home country. And so, yeah, there's there's materials here for everyone to use and it's readily available, and best of all, it's free. So, you know, uh we invite folks to um get access to these materials. And for those that may not speak languages other than English, uh, and we also have a public evaluation tool there to let us know how it's currently doing. Uh, you know, yes, we take those linguistic metrics into account, but we also take uh, you know, the best metric of all, and that's the one that comes from the human being. And so if you guys are telling us that the translation is not doing so well, then we will go back there and retrain the model to make sure that it is.

SPEAKER_00 35:01

It's amazing, it really is. Um, and I think that a lot of the way that you're approaching this, taking into consideration the human being at the end of the cycle is just so important and thinking also about the policies that are put into place, always considering the end user. Um so as you think about your research agenda thus far and the things you've been discovering, what are some policy areas that you would like to see implemented? You can think at the local level, you can think at the national level, you know. Um, like what do you think are our first, second, and third steps that we need to take to help to make alerts and warnings more accessible to this population, this growing population?

SPEAKER_04 35:50

You know, I yeah, this growing population indeed, you know, over 68.7 million Americans speak a language other than uh than English at home right now. And that number is only expected to boom in the next couple of decades. And, you know, let's start off federally, because that's kind of where I've been doing a lot of work with NOAA. And I, you know, I think, you know, there have been recent executive orders that have made English the official language, uh, but there have been several paragraphs within that executive order that state that government agencies are still allowed to communicate to multilingual communities across the country. And, you know, even recently the FCC required WIAs to implement template-based uh multilingual WIAs uh templates from across the board in just a couple of years. So I think that that that policies like those need to continue to be advanced, but we should be able to approach this carefully too. Are we going to approach it in a template format? Are we going to, you know, is this template going to only address one dialect over others? There's so many things that we have to take into consideration that it's not just an easy copy-paste approach as well. And really thinking about that. And I think fundamentally we need to include those experts in the room. Just as much as we have been embracing warning scholars like ourselves over the last couple of years into this space, I think there's should also be space for those that have that specialty. You know, as much as I love to do bilingual weather communication research, I myself am not a sociolinguist. And I collaborate with them all the time because I continue to learn so many new and unique things. And so in the future, I'd love for NOAA to have that similar sort of approach that Environment Canada does in the way that just like they have a 24-7 forecaster desk, they also have a 24-7 translator desk. Let's invest in those resources. Um, because sure, this automation that can get us pretty close to the finish line. But as we were just mentioning, those human beings that have that expertise are considered to be the final mile. Because again, I know much of the meteorology, but there's certified translators out there that that certification exam for translation is among one of the hardest exams to take. And so we have folks that are qualified to do this. And I believe they should also have a seat at the table. I think more locally and statewide too. I think it's, you know, very much a cornerstone and really reimagining how we look at community partners in the first place. Uh, one of the studies that I did a while back was that during the December 10th to 11th, 2021 tornado outbreak, where a lot of immigrant and Latino communities were affected. Uh, agencies like the National Weather Service often think about their core partners being like broadcast meteorologists and emergency managers, et cetera, et cetera. But a lot of the people that really were key to getting a hold of multilingual populations were local community leaders, like local pastors. Uh, that one of them even had a WhatsApp message of like 80 people in their congregation, and he was just putting stuff into Google Translate in real time because he didn't know what else to do. And integrating these folks well beforehand and making them part of that core team, I think, is so essential looking on into the future as we start reaching and engaging more populations. And of course, you know, setting these sorts of standards going forward, you know, investing in an official uh weather and climate dictionary and beyond that, too, let's put it to other hazards as well. And let's um overall create these policies and these resources so that these kinds of efforts could be successful. Because with without those kinds of resources, it, you know, I don't see this uh being a sustainable way to or a really long-term uh way of addressing it, uh at least the current way, of just having volunteer grassroots folks doing a lot of of the translate work because there are certified professionals out there, there are a lot of resources out there, we just really need to invest in them in the future.

SPEAKER_00 39:29

Yeah, and you mentioned other hazards too, and that was a question I I should have asked earlier, but you know, NOAA and the National Weather Service has done so much work in this space and the weather space. What about some of the other agencies? You know, that's one of your policies suggestions about adding other hazards. So where's the where where can other agencies help?

SPEAKER_04 39:50

Yeah, no, I think beyond just weather, there's so many, especially, you know, let me focus more on the WIA perspective, you know, like what will let's think about considering translating like Amber Alerts, for example. That's a whole complete different domain. And I do believe that there's there are various different agencies out there that probably haven't considered this uh initially that would very much would be critical in advancing some of this multilingual risk communication uh across the board. Like I didn't even know that the word debris didn't have a translation in it. I can't imagine uh, you know, that, or I can't imagine that all these other kinds of hazards will have similar kinds of roadblocks along their sort of ways, too. So I would love to expand into uh those uh sort of centers like National Center for Missing and Exploited Children, or uh, you know, even for nuclear hazards as well. Just a lot of things that a lot of immigrant populations might not be as familiar with, I think also deserves a seat at the table here. You know, folks like even NASA, I know, with uh, and then even the national um center for space weather here too, like with all the auroras happening. There's so much yet to be addressed in this, and let alone in Spanish and let alone all the other kinds of languages. And so I I think I really do think there should be some sort of a body up there nationally that kind of is able to bridge all these uh agencies together and look at this multilingual uh aspect more holistically because yeah, we've been advancing a lot of cool stuff within the weather service. I just wish there was a way to make sure that all the other agencies too had a hold of this information as you know, with an increasingly uh changing climate, I think it really does take a collective effort to address this across the board.

SPEAKER_00 41:26

What would be the translation for a geomagnetic storm into Spanish?

SPEAKER_04 41:32

Well, oh my goodness.

SPEAKER_06 41:34

Uh like geomagnetic storm. Anybody listening, put it in the comments. I want to see. I want to see what you're gonna say, but then we're gonna have have the doctor do it. Just go for it, right? What could go wrong? I know.

SPEAKER_04 41:51

Yeah, I was I was gonna go with Tormenta geomagnetica, but then like that's just so weird. Uh like again, there's not really a weird way, uh, uh a concrete way of saying it. And another recent one, too, that actually did a little bit more research on was atmospheric rivers out on the West Coast. Right. Um, cool, that's translated into English, but when you translate that into Spanish, it's literally like a river in the sky. So when we I asked a nationwide sample of Spanish speakers what they associated an atmospheric river with, a lot of people put strong winds rather than floods because they thought it was a river in the sky that was blowing a lot of wind around rather than torrential rainfall. And so it's those kinds of examples where like even terms like in space weather, like, you know, even in English, I have a lot of I'm still learning a lot about that community and like what is a G G3 geomagnetic storm, let alone translate that into another language. Like, and so there's, you know, I I I know this brings us back to the conversation of trying to reduce jargon as much as possible, but I can't emphasize it enough because when we think about all the extremes that happen out there beyond just weather too, there's just so much room for confusion. And we can actively reduce this if we commit ourselves to reducing jargon across the board into other languages. So something for our English speakers to consider as they continue advancing those policies forward, too.

SPEAKER_00 43:03

Yeah, absolutely. I think on every training that Eddie and I go into, we talk about the importance of plain language and simplifying our words because not only does it help English readers and speakers to understand, but it does make it much easier to translate into things that Spanish and other language readers and speakers can read. Um just simplifying. We get so caught up in our own language. And and I mean, I just thinking about putting something like a geomagnetic storm into a 90-character message, you can't do that.

SPEAKER_02 43:40

Yeah, you can't.

SPEAKER_00 43:42

Like it has to be some other way of describing the impacts of that hazard instead of using the scientific language.

SPEAKER_04 43:51

Yeah, I know. And taking into account the cult uh the cultural context too. Like actually, here locally in Champaign Urbana, I've had the honor to work with various indigenous populations as well. And for the first time, I've been learning so much about these languages. One in particular, you know, was the language in Quiche and Kanjobal as well. Kanjobal is more of a Mayan derived language there. Where to them, a lot of uh, you know, when I asked them like what are weather hazards that concern you the most, to them, it's actually sun showers. When whenever it's sunny outside but it's still raining, because to them, culturally speaking, if that were to happen, they would have to go inside their home and close out every single blind possible. And if a raindrop were to fall into your head, you would get gray hair nearly instantly. So they were just like, we don't care about tornadoes. We don't care about hurricanes. We don't care about any of that. I want you to forecast me sun showers. I'm like interested, you know? And it's these sorts of like things that we I'm learning just so much every day. And it's just been a huge privilege to uh learn from these cultures. And how can we make sure not to like erase that? That is something that's respectable and should uh be upheld. But you know use that to also engage them more into the broader aspects of the atmosphere. We can learn so much from these cultures as well and to really make a more accessible warning communication system for us. But yeah, they learned so much about it recently and it's just been it blew my mind.

SPEAKER_07 45:13

Yeah yeah well I remember reading one in English and it was somebody who's trying to be creative and I mean it's almost a marketing game at some point and they said you know weather is knocking at your door be prepared right and it's examples like that that could you imagine translating that into other languages that weather is now knocking at your door and what does that mean? And and it's just uh again example we use plain language folks that's what we need to use be very clear avoid jargon and and all of that. Now for my question I think you've answered it like five times already because it's really what are you going to work on next and where do you see your research which I will give you that that time um but from hearing everything that you've said and and I really hope that everybody who's listening is trying to take notes and say okay there's a lot of resources out there. So I want to give you a chance and you can think about it for a second talk directly to those listening and basically think about it from the the lens of someone who is in a small agency who has a diverse population and is really trying their best and they are probably in over their head but it's like if they were going to ask you can you give me like three things that I can either do or go see that I can read that I can use to make my messaging more effective number one identify you know the local community leaders that resonate with that given population.

SPEAKER_04 46:53

It goes so far. You know they actually decide a lot of the time to step up during these kind of extreme weather events and provide some help along the way we saw that a lot in the aftermath of that tornado outbreak that I was speaking to you about before they were really hungry for a lot of the weather and climate risk communication that was happening at the time and they wanted to translate it. And even in the aftermath of that study, as a result, we connected the local community leaders with the with the agencies and with the emergency managers of the area and they ended up collaborating on something so special that um people that from the refugee or local refugee organization there at the time actually created interactive videos that every time a new immigrant would come into the country, they would have videos from them from the people themselves describing how they experienced that tornado. And so with limited resources there are community members out there that are just so eager to help and identifying those local community leaders I think would be a very suitable first step in engaging these communities because not only does it is it effective but it reaches them really directly there. Number two would be to look at these resources that exist present day. Yeah this uh AI translation program we tried to get the word out there so much on it um but yeah there's readily available materials I know it's difficult with limited resources to uh ask someone to translate into a complete other language make sure it's right but you know we're bringing out these resources um you know even on my website I have a website full of things of like multilingual resources that folks can access. And so to make things easier on there, we'd most definitely recommend folks to become more familiar with this. And I'm not asking like a real live translation during some of the biggest events but you know even right before something were to happen, you know, getting even like one or two posts in Spanish I've seen in social media that it also resonates a lot with folks because at the end of the day we're a community and we want to help each other out. And you know that I've seen especially with like shares and interaction people love to share that out to their community members uh as well. And then number three I think it's also just to to sick uh take a whole step back here. Now I feel like I when I got in my when I did my PhD in Oklahoma I saw all the broadcast meteorologists there um where they were just you know they knew that Oklahoma was tornado savvy and they were going well above and beyond with the some of their terminology that they were using. They even show like correlation coefficient on TV and we're like whoa um and it's yeah it's to acknowledge that a lot of these people may not be familiar with these hazards in the first place. And you might be in an area you know that's like experienced floods all the time. Maybe there's a significant one that people still remember to that day. I think number three is to just take a step back and realize that that may not be necessarily the case for these communities. It's okay to continue to post those educational campaigns over and over because there are people out there that really do need it. And to not assume that that area in general is is aware of that hazard altogether. Like one example I always like to share is that in the word of among Arabic speakers, the word tornado and hurricane are the same word. It's like a spinning cyclone of wind. So there's not actually a way to describe a tornado in that language. And so always taking that step back sharing those educational campaigns I just think is so critical and making sure that everyone starts on the same foot so that we can end in the same place as well. So yeah, find that local community leader that um is willing to work with you use those existing resources and realize that this community may be small and mighty but they also don't may not have these kinds of experiences. So as you engage them along the way just have an open mind and you know always there's always these resources there to help along the way as well.

SPEAKER_07 50:39

I love that.

SPEAKER_04 50:40

And then can you just say your website can you put it out there so we can link it oh yeah of course uh my own website is uh josefdrujiofalcon.com but uh we have there's a multilingual resources uh page there where I put all the existing semi-dictionaries everyone has a there's not an official dictionary out there but there for example there's graphics out there there's different campaigns by FEMA and the weather service that they've done in the past so you can access that there too. Now I've even, as a nerd myself have been trying to find all the existing dictionaries. So like I have some from like the the the weather service in Mexico and Uruguay too just to see how they define it. So there you go. But yeah and then uh I'll also link my lab's website uh publish.illinois.edu slash alertas lab. I'm part of the advancing language equity and resilience in the Asturix Science or Alertas Lab. We're we're continuing work in this area and you can keep up with all our recent publications and research and uh you know uh different kinds of articles that we've been we've been advancing in this area and it's moving quick with this evolution of AI and all these languages hitting us at once I always tell folks here in the United States we practice over 150 languages uh yet our emergency communication system is only relying on one now in terms of my job here I think I have so much work to do over the I think this is a full on career you know I even by the end I don't think I'll ever get anywhere close to those 150 but hey man can dream and so I will keep dreaming and make sure that we can make it more accessible for folks along along the way and Joseph what are you working on next yeah uh right now actually it's a lot of the validation work related to AI translation uh so uh we actually are working with sociolinguists and develop metrics for accuracy readability uh satisfaction and oh my gosh I I'm totally blanking on the last one but it had things related to do with syntax with the actual grammatical structure of Spanish but then also the warning response aspects to it so we asked we included some things from the warning response model there and we uh on our first test we got a Spanish translation of a weather service meteorologist how it's traditionally done and then the AI translation um and we uh analyzed we randomly or we sent out the survey and people were split half half when which uh product they saw and they were able to evaluate it with those different metrics uh within the AI translation program. Uh so not to spoil it too much but that was the first study I've ever been excited about not having a statistically significant result because it resonated there was no statistical difference between the two um for that one hurricane product though. So let me be caveat there. But even then we spent over five years training that to where hurricanes were able to reach a level to where people feel just as comfortable and actually there was a small statistical significance within the AI product that were actually even more willing uh their self-efficacy their uh motivation to take protective action was actually even slightly higher and so again it's only one product in one scenario and it's also simulated so you can only uh take it with a grain of salt but I you know I'm gonna continue evaluating some of these various kinds of products in different contexts as well because hurricanes for the most part are more familiar in Latin America. So you know we'll try it for flash floods next year or wildfire threat to see how that holds I'm also really excited to evaluate it this upcoming year in our survey where we put in a chat GPT translated text versus one from the AI translation program versus one that has been made and we're going to evaluate some of that too. And so a lot of that you know I'm really centering my future work is really centering on evaluation. What makes an effective warning message in multilingual settings within this context of AI and how can we move to make sure that we're doing it in an ethical and in a way that is accessible to folks. So we're centering in on that program and then operationally with the AI translation program. You know we as I mentioned before we've been putting out products for the World Cup this summer but um we're actually running a test bud experiment soon to see how it would look like in practice. I mean we're already kind of doing it in practice but just seeing how it could look like because we're gonna apply some of these same lessons and do it all over again for the 2028 28 LA Olympics uh because multiple international speaking audiences will be there too so we're advancing some of our resources in that format and my my my dream I'm knocking on wood here would be to go to the Olympics and interview the international tourist community and kind of get their perspectives on this on like you know we've really reframed this program to not just benefit the public but international tourists that bring so much uh money to our local American economy each year and I want to get their insights too you know what are we missing from that perspective? So all in all some very fun research uh in the future and I have a really awesome team here of graduate students that have just been so awesome and uh yeah I'm looking forward to continuing to uh to advance this area in the future.

SPEAKER_00 55:42

So that's very exciting. So when we send you off to spend time at the Olympics with your graduate research team which sports are you gonna go to because you know they're gonna have to talk about their sport in their language.

SPEAKER_04 55:55

I know see I I selfishly want to go to the gymnastics competitions. I'm just like come on like um actually the the last American Meteorological society meeting we were like in the third floor of the convention center in Houston and in the first floor there was a Simone Biles camp for for like I think it was like for girls. And so Simone Biles was literally in the same hotel as us and I'm just like this is a sign that I we have to like I don't know if she'll be at the 28 Olympics but yeah that um you know so that's the number one sport I I'm I'm watching to go to but there's so many out there too and yeah I'm excited to see I don't even know how I designed a study like that in the first place.

SPEAKER_00 56:34

So I'm good thing is I have time but you have time and I'm sure that there's going to be people who want to contribute to something like that. So you can form a whole research team to to get out there and do that. No, it'd be so much fun. You'll have to join us too great place. I I was living in California when they had their last Olympics in Los Angeles a long time ago I think it was 80 80 which which what year was it when they last had it yeah I don't know Jeanette you're dating yourself I I don't know but we could all be there potentially for the next one so that'll be the fun thing. 1984 looks like 84 okay thank you yeah shows you how old I am aging well like I were uh thank you Joseph this has been really interesting enlightening you've given us so many resources that people should be turning to to get more familiar with I love that you've shared with us where things are going and to me I think it's a really positive thing that the AI models that the National Weather Service has been helping to train are actually reaching a similar level of accuracy and readability and trustworthiness that a human being is because that's the direction things are headed, whether we like it or not. And if if there are AI systems out there that are able to do this effectively in English and in translation, then it could improve our messaging.

SPEAKER_04 58:28

It really it really could because you know before it we would have never imagined trying to scale up to so many languages before you know when I first started this journey I would have never imagined that a decade later we'd have 12 languages that we're actively translating in this AI model. You know there's I I am I'm a fond believer that yes, uh we have to do it right and ethically and make sure it it engages a human being but there is so much potential here to expand especially in the multilingual world. So we're investing in it and just going to keep working through this and I do hope it is a scalable model. You know we're uh to our friends at out in FEMA and other government agencies this could be applied in those contexts as well so we're just you know part of my job is kind of just like taking notes on everything we're doing along the way and putting this work out there like with this AI piece that we just published and that really is just a long publication of every step we took along the way because we're trying to be uh uh to tell other agencies like hey you can implement this too and here's a how-to guide so you can avoid all the mistakes we made along the way so fingers crossed for that yeah well I just appreciate it because I like Jeanette said it's important to maintain optimism and hope because choosing not to do anything because you're frustrated and you don't know how is not the right choice.

SPEAKER_07 59:45

You need to do what you can and if all you can do is is just that little bit then do that little bit and then the next time try to do a little bit more and just keep advancing and progressing and realize you're not alone that you have resources just like you mentioned on there and again we're gonna try to get all the links um to this so that people listening can go to it because you shouldn't have to battle this by yourself out there. And unfortunately there are people that think they do have to but I'm telling you you don't have to you you can rely on amazing research you can look at other practitioners there are many examples out there. And again just thank you for your research and and don't stop and you know just do what you can to make it go faster, right?

SPEAKER_04 1:00:31

Yeah yeah no oh my goodness I wish yeah no for sure and I I do want to end it on a sweet sweet note here on a very full circle moment here because like I not only have been doing this work but I I still have a broadcast bug and so I've locally been doing weather communication here locally through a Facebook page it's completely in Spanish and the other day my mom uh and I were going to our favorite taco truck in the area and um as we're ordering our our our order there the the lady that's currently making our food looks at me she's just like oh my God hola señor metrona and she like look goes to my mom and she's just like that he helped us out a lot this because we just broke records for tornadoes out here in Illinois and she was just like you know like he really helped us out our local business we just want to thank y'all so much. We're gonna put extra chicken in your burrito and like my mom looked at me and that was a full circle moment there of like her dragging me out there and then hey bare minimum it at least results in a bigger chicken burrito that day around but she was so proud and at the end of the day that's that's what we do it for for the community. And those those are the moments that keep my energy up despite all everything that's been going on and you know because it's they're worth it at the end of the day. And I think our field this is why I admire the emergency management and risk communication field so much. We're all so devoted to our communities and that's what we've got to keep working for. So there you go. That's a fun story to end it.

SPEAKER_00 1:01:54

Well thank you Joseph and this has just been a really wonderful conversation. I'm so happy that you were able to join us and I know our listeners are getting a lot out of this conversation today. So those of us those of you who are listeners to the Alert or watchers of the alerting authority please do check out the resources that were shared. Those are available to you freely um which is just amazing with our uh limited resources these days um and also if you have any questions or needs related to your alerts and warnings please remember that Eddie and I are available to chat with you at any time. We would love to review your messages, your templates, help to provide guidance about your policies. And we also spend some time traveling to other states and talking about how to improve your alert and warning practices from top to bottom through half day and day long trainings and they can be fun but and there's a lot of information to share and we're happy to do that. So please reach out to us at any point in time.

SPEAKER_07 1:03:02

You can find us on the Alerting authority and thewarnroom.com I just want to say another thank you to everybody listening like Jeanette and I don't talk about this very much and and Joseph you get to be a part of this but um our our podcast on Apple and on Spotify um like we've found out we are the number one podcast for public safety alerts the number one for wireless messaging the number one for alerting skills and and there's some other different categories where we're either number one or we're in you know the top four for those and it's like just thank you because we want to make sure this is what you want to hear and the fact that we can connect to amazing people like the one we have today um I think is just going to help everybody. So thank you for listening. Thank you for hitting the like button the subscribe button for leaving um comments for leaving reviews that that does make a difference and it allows us to continue to do this. So oh I guess it's my turn to say and this is another episode of the Alerting authority and we know that every second has a story and we're grateful to be able to help share yours.

SPEAKER_05 1:04:19

And that's it