| 3D | Doctoral School "Dynamiques du Developpement Durable", University of Toamasina |
| APA | American Psychological Association (citation style) |
| API | Application Programming Interface |
| AR | Autoregressive (decoder family) |
| ASR | Automatic Speech Recognition |
| AUC | Area Under the Curve |
| BLEU | Bilingual Evaluation Understudy (translation metric) |
| BPE | Byte-Pair Encoding |
| CER | Character Error Rate |
| CTC | Connectionist Temporal Classification |
| CV | Common Voice (Mozilla multilingual speech corpus) |
| CV-MG | Common Voice, Malagasy subset |
| DNN | Deep Neural Network |
| DOI | Digital Object Identifier |
| ECE | Expected Calibration Error |
| EDTM | Ecole Doctorale Themes Multiples (here EDTM 3D) |
| ELBO | Evidence Lower Bound |
| EM | Expectation-Maximization algorithm |
| F1 | Harmonic mean of precision and recall |
| FLEURS | Few-shot Learning Evaluation of Universal Representations of Speech |
| FLEURS-MG | FLEURS benchmark, Malagasy subset |
| FLOPS | Floating-Point Operations per Second |
| FR | French (code-switching language tag) |
| GMM | Gaussian Mixture Model |
| GPU | Graphics Processing Unit |
| HMM | Hidden Markov Model |
| IMRaD | Introduction, Materials and Methods, Results, Discussion |
| IPA | International Phonetic Alphabet |
| ITP | Iterative Pseudo-Labeling |
| JS | Jensen-Shannon (divergence) |
| JSON | JavaScript Object Notation |
| KL | Kullback-Leibler (divergence) |
| LAS | Listen, Attend and Spell |
| LF-MMI | Lattice-Free Maximum Mutual Information |
| LLM | Large Language Model |
| LM | Language Model |
| LoRA | Low-Rank Adaptation |
| MAP | Maximum A Posteriori |
| MFCC | Mel-Frequency Cepstral Coefficients |
| MG | Malagasy (language code) |
| MLE | Maximum Likelihood Estimation |
| MMI | Maximum Mutual Information |
| MMS | Massively Multilingual Speech |
| MoE | Mixture of Experts |
| NIST | National Institute of Standards and Technology |
| NLP | Natural Language Processing |
| NMT | Neural Machine Translation |
| OWSM | Open Whisper-style Speech Model |
| PLP | Perceptual Linear Prediction |
| RNN | Recurrent Neural Network |
| RNN-T | Recurrent Neural Network Transducer |
| ROC | Receiver Operating Characteristic |
| SIGUL | ACL Special Interest Group on Under-resourced Languages |
| SI-SNR | Scale-Invariant Signal-to-Noise Ratio |
| slimIPL | Slim Iterative Pseudo-Labeling |
| SNR | Signal-to-Noise Ratio |
| SSL | Self-Supervised Learning |
| TASLP | IEEE/ACM Transactions on Audio, Speech and Language Processing |
| TDNN | Time-Delay Neural Network |
| TDNN-F | Factorized Time-Delay Neural Network |
| TPU | Tensor Processing Unit |
| USM | Universal Speech Model |
| VAD | Voice Activity Detection |
| VAE | Variational Autoencoder |
| WavLM | Speech pre-training framework with denoising |
| WER | Word Error Rate |
| WFST | Weighted Finite-State Transducer |
| XLS-R | Cross-Lingual Speech Representation (large-scale) |
| XLSR | Cross-Lingual Speech Representation (initial release) |