ASR

Robustness / Generalization

Title	Year	Author	Link	Memo
TOWARD DOMAIN-INVARIANT SPEECH RECOGNITION VIA LARGE SCALE TRAINING	2018	Arun Narayanan et al.	pdf	a single domain-invariant model for varied use-cases from multiple domains
RETHINKING EVALUATION IN ASR: ARE OUR MODELS ROBUST ENOUGH?	2021	Tatiana Likhomanenko et al.	pdf	evaluation across multiple test sets is necessary (at least tedlium)

Title	Year	Author	Link	Memo
EFFICIENT KNOWLEDGE DISTILLATION FOR RNN-TRANSDUCER MODELS	2020	Sankaran Panchapagesan et al	pdf	use only target label and blank label for distill rnnt

Title	Year	Author	Link	Memo
CROSS-UTTERANCE ASR RESCORING WITH GRAPH-BASED LABEL PROPAGATION	2023 ICASSP	Srinath Tankasala et al.	pdf	rescore a set of utterances together using label propagation over acoustic similarity graph

Title	Year	Author	Link	Memo
SPEECH REPRESENTATION LEARNING THROUGH SELF-SUPERVISED PRETRAINING AND MULTI-TASK FINETUNING	2021	Yi-Chen Chen et al	pdf	multitasking on top of SSL

Title	Year	Author	Link	Memo
HIERARCHICAL SOFTMAX FOR END-TO-END LOW-RESOURCE MULTILINGUAL
SPEECH RECOGNITION	icassp 2023	Qianying Liu et al	pdf	applied hierarchical softmax to cross lingual tokens