avatar

Adel Moumen

PHD student
SpeechBrain Core Maintainer

University of Cambridge
am3303@cam.ac.uk


About Me

I am a 22-year-old PHD student at the University of Cambridge under the supervision of Prof. Phil Woodland. I completed my Bachelor’s and Master’s degree in computer science and AI with distinction in an innovation and research-devoted curriculum and earned a two-year entrepreneurship diploma in 2022. I am professionally also contributing to the development of SpeechBrain, an all-in-one, open-source, PyTorch-based speech processing toolkit with more than 8,800+ stars on GitHub. At SpeechBrain, I lead the core efforts of the toolkit. In 2019, I started as an autodidact on deep learning and helped frame the largest French AI community.

Research Interests

My current research interests revolve around improving Deep Neural Networks. Specifically, I focus on exploring innovative methods to enhance their efficiency by rethinking their core architecture. I am particularly interested in addressing concrete challenges, such as the gradient exploding problem in Recurrent Neural Networks. My research primarily applies these concepts to the field of Automatic Speech Recognition.

SpeechBrain

I serve as a core maintainer of the SpeechBrain toolkit, responsible for its core development and overall management. My role entails actively supporting the toolkit by engaging in discussions, addressing issues, and reviewing pull requests. Additionally, I focus on expanding the toolkit’s capabilities by introducing new features for automatic speech recognition.

One of my notable contributions was integrating the openAI’s Whisper model. My ongoing work involves incorporating advanced decoding methods into SpeechBrain speech recognition systems. It includes integrating CTC frame-synchronous beam search and CTC/Att joint decoding, leveraging language models such as kenLM and TransformerLM (e.g., GPT2) for improved performance. I am also working on a new large-scale ASR model, “openSB-ASR,” trained solely on open-source data.

News

Publications

  1. Adel Moumen, Titouan Parcollet
    IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023.