User profiles for Scott Novotney
Scott NovotneySenior Applied Scientist, Amazon Verified email at amazon.com Cited by 441 |
[PDF][PDF] Cheap, fast and good enough: Automatic speech recognition with non-expert transcription
S Novotney, C Callison-Burch - … of the North American Chapter of …, 2010 - aclanthology.org
Deploying an automatic speech recognition system with reasonable performance requires
expensive and time-consuming in-domain transcription. Previous work demonstrated that non-…
expensive and time-consuming in-domain transcription. Previous work demonstrated that non-…
Unsupervised acoustic and language model training with small amounts of labelled data
S Novotney, R Schwartz, J Ma - 2009 IEEE International …, 2009 - ieeexplore.ieee.org
We measure the effects of a weak language model, estimated from as little as 100k words of
text, on unsupervised acoustic model training and then explore the best method of using …
text, on unsupervised acoustic model training and then explore the best method of using …
[HTML][HTML] Large-scale design and refinement of stable proteins using sequence-only models
JM Singer, S Novotney, D Strickland, HK Haddox… - PloS one, 2022 - journals.plos.org
Engineered proteins generally must possess a stable structure in order to achieve their
designed function. Stable designs, however, are astronomically rare within the space of all …
designed function. Stable designs, however, are astronomically rare within the space of all …
[PDF][PDF] Analysis of low-resource acoustic model self-training
S Novotney, R Schwartz - Tenth annual conference of the …, 2009 - isca-archive.org
Previous work on self-training of acoustic models using unlabeled data reported significant
reductions in WER assuming a large phonetic dictionary was available. We now assume only …
reductions in WER assuming a large phonetic dictionary was available. We now assume only …
[PDF][PDF] Crowdsourced accessibility: Elicitation of Wikipedia articles
S Novotney, C Callison-Burch - Proceedings of the NAACL HLT …, 2010 - aclanthology.org
Mechanical Turk is useful for generating complex speech resources like conversational
speech transcription. In this work, we explore the next step of eliciting narrations of Wikipedia …
speech transcription. In this work, we explore the next step of eliciting narrations of Wikipedia …
Cue vectors: Modular training of language models conditioned on diverse contextual signals
S Novotney, S Mukherjee, Z Ahmed… - arXiv preprint arXiv …, 2022 - arxiv.org
We propose a framework to modularize the training of neural language models that use diverse
forms of sentence-external context (including metadata) by eliminating the need to jointly …
forms of sentence-external context (including metadata) by eliminating the need to jointly …
Improving accuracy of rare words for rnn-transducer through unigram shallow fusion
End-to-end automatic speech recognition (ASR) systems, such as recurrent neural network
transducer (RNN-T), have become popular, but rare word remains a challenge. In this paper, …
transducer (RNN-T), have become popular, but rare word remains a challenge. In this paper, …
Attention-based contextual language model adaptation for speech recognition
Language modeling (LM) for automatic speech recognition (ASR) does not usually incorporate
utterance level contextual information. For some domains like voice assistants, however, …
utterance level contextual information. For some domains like voice assistants, however, …
Robust keystroke transcription from the acoustic side-channel
D Slater, S Novotney, J Moore, S Morgan… - Proceedings of the 35th …, 2019 - dl.acm.org
The acoustic emanations from keyboards provide a side-channel attack from which an
attacker can recover sensitive user information, such as passwords and personally identifiable …
attacker can recover sensitive user information, such as passwords and personally identifiable …
[PDF][PDF] Unsupervised Arabic Dialect Adaptation with Self-Training.
S Novotney, RM Schwartz, S Khudanpur - InterSpeech, 2011 - researchgate.net
Useful training data for automatic speech recognition systems of colloquial speech is usually
limited to expensive in-domain transcription. Broadcast news is an appealing source of …
limited to expensive in-domain transcription. Broadcast news is an appealing source of …