 Our next presenter is Karla Viegas from the Language Technologies Institute. Our title is Machine Learning Frameworks for Stuttering Prediction. This man on the picture is Lewis Carroll. He's the author of Alice in Wonderland, this beautiful story that most of us know. However, I think probably few know that his initial dream was to become a priest. He had to give up this dream because of a speech impediment. Lewis Carroll was a stutterer. Giving up a dream because of stuttering, why would you do that? Most of us have no idea how much of a burden it can be to live a life with stuttering. Stutters feel ashamed because they're not able to speak as everybody else. They encounter incomprehension from others and are often viewed as less intelligent, although that's not true. You might think that giving this talk here or having a job interview is challenging, but for stutters even ordering food can be challenging. There is no cure for stuttering. There are only techniques that can be learned to speak more fluently. And there are 3 million Americans who stutter. The causes are not known, but the probable factors are genetics or social and family environments. However, what we know is that there is a close relationship between emotions and stuttering. Self-assessment studies reveal that if stutters have a negative attitude towards the own stuttering, stuttering severity is increased. So this was studied with the self-assessment studies, but it was not studied with objective measurements, which is what I want to tackle during my PhD. So my initial goal is to measure the relationship of emotions and the stuttering frequency using machine learning techniques. For this, I designed an experiment in which I will expose stutters to different levels of stress and also different emotions. During this experiment, I will be recording audio, video and physiological signals. Those will then be analyzed by two different frameworks, as you can see here. One to detect the emotion and the other one to detect the stuttering severity. The next step would then be to see if it is even possible to predict stuttering. This could be extremely cool if combined with a biofeedback system, which we can imagine to be, for example, a smartwatch. So through the smartwatch, we could measure all day long physiological signals and also analyze the audio using machine learning models. This combined with a biofeedback system that can be, for example, a vibration could remind the stutterer to use the fluency techniques whenever they are about to stutter. This has been so far quite challenging for me because there's not much data available and also not much prior work approaching this problem from the computational perspective. However, the outcome that can change the life quality of stutterers is so significant that this keeps me motivated and also continuing, even though there's not much resources, because it can give freedom to stutterers to follow their dream, even if they stutter. Thank you very much.