A hub for your training needs

Anchalee Phataralaoha: Welcome to our technology conversations podcast, where we discuss technology related topics from how to find resources for your technology needs to how technology can impact our lives. My name is Anchalee Phataralaoha, and I will be your host.

With us today is Dr. Ratree Wayland, a professor of linguistics here at the University. So welcome.

Ratree Wayland: Thank you, thank you for having me.

Anchalee Phataralaoha: Yes, and thank you for being here. To start off, could you share with us your journey to where you are today?

Ratree Wayland: I was born and grew up in northeast in Thailand and I was educated up until the undergraduate level in Thailand. And I obtained my master's and PhD in linguistics from Cornell University in 1997.  And then after that, I assume a NIH postdoc position at the University of Alabama Birmingham up until 1999 when I started the assistant professorship in linguistics at UF.

Anchalee Phataralaoha: So you have been here for quite a long time?

Ratree Wayland: A long time, yes, since August 1999. So that's what, 26, 27 years?

Anchalee Phataralaoha: Yeah, almost 30 years.

Ratree Wayland: Yep, has been here and has never been anywhere else.

Anchalee Phataralaoha: So what do you teach and what is your research interest?

Ratree Wayland: My main teaching responsibility is I teach all phonetics related courses for the department. So I teach one core required course at the undergraduate level called the Sounds of Human Language. And then I teach the speech acoustics courses, two of them in sequence, one is called the Fundamental Phonetics and I think the other one is Advanced Phonetics.

Anchalee Phataralaoha: And what about your research?

Ratree Wayland: My research, I started my career looking into how speech patterns of people who learn a second language differ from those of the native language because I myself have been a lifelong I can say second language learners of multiple languages. So I have been always interested in how we produce the second language sounds differently from the native speakers and how we can improve our production of the second language to enhance our communication with native business. So that's how I started. And I have still..I should say, I'm still working on that kind of research. But more and more recently, meaning the past may be five years or so, I have become more and more interested in looking into speech as a what we call biomarker of diseases or conditions caused by, say, neural developmental deficits or normal motor control problems such as Parkinson's disease, ALS, and so on and so forth.

Anchalee Phataralaoha: So when you do your research, do you need particular tools or any device to detect those things?

Ratree Wayland: Yes. Originally, at the very least, we have to have software to analyze speech patterns, what we call speech acoustic analysis. So we have to have that as a basic. So it is a very time-consuming, arduous process where you take the audio recordings of speech, you know, be it speech from a second language learner or speech from clinical populations and what not. So the first thing that we do, we have to basically analyze speech for its acoustic properties. And we have to rely on speech acoustic analysis software for that. So it is a very time -consuming. It can take days and months to analyze maybe an hour of speech recordings. But these days, we have more advanced tools that allow us to complete our job a little bit more; I would say definitely faster but probably more accurately in some fashions, but definitely these modern tools have been time-saving devices in our research.

Anchalee Phataralaoha: Does that include AI because these days everyone talks about AI doing this and doing that?

Ratree Wayland: Oh yeah, yes definitely big time that involved AI so we have tools such as what we call the Forced Aligner, which is the machine learning tools that allow us to transcribe speech, you know, transcribe the audio or match the text to audio. But of course, it's, you know, it depends on how the tools and the machine or the aligner has been trained. The accuracy would vary depending on, again, the amount of training the machine has gone through. So it still requires manual checking for accuracy, but it substantially reduces the amount of time because we don't have to do everything from scratch. We simply just go through and make sure that everything is aligned properly and correctly.

Anchalee Phataralaoha: Any particular AI model tools that you use?

Ratree Wayland: We, for this part, for the Aligner or Forced Aligner, basically the matchup between the text and the audio we have been using what we call Montreal Forced Aligner. So that has been the tool that we have been using. As far as speech acoustic analysis, we have been using all kinds of automated software, but also we are leaning more recently to what we call self-supervised machine learning models such as Wave2Vec, WaveLM, or other self-supervised learnings that would allow us to basically analyze the speech pattern automatically without actually the transcription. So it bypassed the transcription completely. So it allows us to be able to analyze speech without first, you know, align the audio signals with the text. So it's a very, very useful tool that allows us to analyze speech without having to first transcribe it. and we can analyze speech in a large quantity as well. That takes hours as opposed to months or maybe sometimes a year to go through that.

Anchalee Phataralaoha: So does it mean with the way that machine can help you work faster -- does it mean you can do more research?

Ratree Wayland: Yes, I'm really encouraged by the potential, the productivity aspect of my research because of all the tools that we are employing. So instead of, you know, spending six months analyzing recordings of, let's say, 20 speakers, we can do that in a matter of hours, right? So yeah, so we can actually do more research as a pastor, I mean, I should say as a faster pace, And hopefully more accurately and also more in depth as well, you know, because it comes from the fact that we can analyze a larger amount of speech. So hopefully the results will also be more reliable as well.

Anchalee Phataralaoha: So now if we, you know, going kind of another turn to teaching, do you use any tools, you know, for teaching compared to let's say decades past?

Ratree Wayland: As my teaching goes, I have always incorporated whatever tools that we have in our repertoire to demonstrate things. For example, I have been using, I have been using tools that allow the students to see how the speech spectrograms is generated real-time. So I can demonstrate to them what the difference is acoustically between two-speed sounds live, you know, as in the classroom. So I will produce sounds, different sounds, and I can, you know, show them this is how they are different and so on.

Anchalee Phataralaoha: So since you've been teaching and researching for such a long time, you know, I'm sure you have seen a lot of changes. And these days, things change much faster at a faster pace. So what do you find is the most challenging, you know, in your role?

Ratree Wayland: The most challenging, I would say not only the amount of research that we can do, you know, but I think what's not changing is the, well, how should I say, it is the bridging the gap between what we learn and what people actually need, right? So I think we can do more and more research these days, but I think it remains a challenge whether or not what we have learned through research actually will actually be benefiting the society, the people out there. So a lot of my job is about thinking about translation and interpretation, working with collaborators across disciplines to make sure that the research or the models that we use are not only accurate, but also intuitive, meaningful, and accessible to people outside of the classroom or outside of the academia, right? So it's important to do research in real-world use cases. For example, I always strive to do research that would answer questions such as how can we make speech recognition more inclusive, how we can detect early changes in speech patterns that may signal neurological disorders, or how we do  or measure progress in language learners in a way that would reflect actual communicative abilities. I think with more and more research or at a faster pace that we are gaining all the information, we still need to be mindful of, as I said, bridging the gap between what we know and how that would translate into the actual benefits to people in the real world and to people in the society at large.

Anchalee Phataralaoha: So if someone is interested to learn more about these areas or even your research, where can they find more information or even reach out to you?

Ratree Wayland: They can go to the Linguistic Department website and look me up there and my email should be listed there and they can email me. And I'm always open to collaborative effort with students, with faculty from different departments. So I will be more than happy to discuss what we are doing and what we can potentially do together to learn more about speech itself and also how speech can or on what we know about speech patterns and how can it help us understand normal speaking, listening conditions, but also the speech patterns in among, among clinical populations as well.

Anchalee Phataralaoha: Any final thoughts?

Ratree Wayland: Well, what would be the final thought? I think one thing that I can say is that, you know people don't often think about speech as being important because it's something that we do all the time. But in fact, it is the most human things that we do, right? It's how we connect, it's how we express ourselves and navigate the world. What excites me the most is using AI and new tools that I have just mentioned, not to replace human understanding, but to enhance it. With the right tools we can now see subtle patterns in speech that help us better understand identity, learning, health, and changes over time.

Anchalee Phataralaoha: Thank you very much for sharing your insights and your knowledge.

Ratree Wayland: Well, you're welcome. Happy to be here.

Anchalee Phataralaoha: And that's it everyone. We will see you next time for a topic of interest in IT.