Scholar
Ronan Collobert
Google Scholar ID: 32w7x1cAAAAJ
Machine Learning Research, Apple
Citations & Impact
Citations
28,821
H-index
46
i10-index
86
Publications
20
Co-authors
39
list available
Contact
Publications
Resume (English only)
Academic Achievements
- Published paper in 2021: 'Word Order Does Not Matter For Speech Recognition', proposing a weakly supervised ASR training method without word order in transcripts, achieving near-supervised performance on LibriSpeech
- Developed several influential open-source ML tools:
- - flashlight: an efficient modern C++ ML framework with autograd and deep learning support
- - wav2letter: end-to-end speech recognition toolkit, now part of flashlight
- - Torch7: Lua-based ML library offering Matlab-like environment
- - SENNA: fast NLP tagger for POS, NER, SRL, etc.
- - SVMTorch: early SVM library (no longer maintained)
Background
- Research scientist at Apple, Machine Learning Research
- Research interests include machine learning on structured data (NLP, image, speech)
- Focuses on semi-supervised, weakly supervised, structured, or compositional learning techniques
- Dedicated to large-scale learning and software enabling unconventional machine learning approaches
- Both research and software follow the KISS (Keep It Simple, Stupid) philosophy
