The Icelandic Centre for Language Technology hosts a seminar on September 18th at 12:00, at Edda Conference Room, Arngrímsgata 5, Reykjavík
Shijun Wang, a PhD student in speech recognition gives a talk titled: Deep Learning for Paralinguistic Representation Learning. The talk will be in English.
Paralinguistic elements in speech, such as emotion, speaking speed, and volume, play a crucial role in conveying the speaker's attitudes or intentions. Unfortunately, these elements are often overlooked. In this presentation, innovative approaches are introduced to effectively extract paralinguistic representations and tackle common challenges. Furthermore, these representations can significantly enhance various speech processing tasks, including voice conversion, emotion recognition, and emotional text-to-speech synthesis. In conclusion, this presentation underscores the potential of utilizing paralinguistic representations to enhance the quality of human-computer interactions.