9 Best Articles in 2022
PNASNews
Racial disparities in automated speech recognition
PNASNews
From 2020 · Automated speech recognition (ASR) systems are now used in a variety of applications to convert spoken language to text, from virtual assistants, to closed captioning, to hands-free computing. By analyzing a large corpus of sociolinguistic interviews with white and African American speakers, we demonstrate large racial disparities in the performance of five popular commercial ASR systems. Our results point to hurdles faced by African Americans in using increasingly widespread tools driven by speech recognition technology. More generally, our work illustrates the need to audit emerging machine-learning systems to ensure they are broadly inclusive.
Automated speech recognition (ASR) systems, which use sophisticated machine-learning algorithms to convert spoken language to text, have become increasingly widespread, powering popular virtual assistants, facilitating automated closed captioning, and enabling digital dictation platforms for health care. Over the last several years, the quality of these systems has dramatically improved, due both to advances in deep learning and to the collection of large-scale datasets used to train the systems. There is concern, however, that these tools do not work equally well for all subgroups of the population. Here, we examine the ability of five state-of-the-art ASR systems—developed by Amazon, Apple, Google, IBM, and Microsoft—to transcribe structured interviews conducted with 42 white speakers and 73 black speakers. In total, this corpus spans five US cities and consists of 19.8 h of audio matched on the age and gender of the speaker. We found that all five ASR systems exhibited substantial racial disparities, with an average word error rate (WER) of 0.35 for black speakers compared with 0.19 for white speakers. We trace these disparities to the underlying acoustic models used by the ASR systems as the race gap was equally large on a subset of identical phrases spoken by black and white individuals in our corpus. We conclude by proposing strategies—such as using more diverse training datasets that include African American Vernacular English—to reduce these performance differences and ensure speech recognition technology is inclusive. · Shared by 33, including Stanford NLP Group, Aaron Gustafson, Joanna J Bryson
Mozilla
Announcing the Initial Release of Mozilla’s Open Source Speech Recognition Model and Voice Dataset
Mozilla
6 min read · From 2017 · With the holiday, gift-giving season upon us, many people are about to experience the ease and power of new speech-enabled devices. Technical advancements have fueled the growth of speech interfaces… · Shared by 68, including Marco Unternaehrer, Tibor Martini 🇺🇦, Oleg Baskov, Isabel Drost-Fromm, Sascha Eggenberger, Mara Averick, Victor Lee, Thomas Power, Ferit (at 🏠) 🌙, Charlie O'Keefe, Data Science Renee
The Gradient
New Datasets to Democratize Speech Recognition Technology
The Gradient
~17 min read · 2021-12-14 · MLCommons.org introduces two new public datasets for speech recognition. The People’s Speech is the first large-scale, permissively licensed ASR dataset that includes diverse speech and environments. · Shared by 34, including AI, Andreas Staub, Brian Ahier, ipfconline, Mike Tamir, PhD
microsoft.com
Microsoft researchers achieve new conversational speech recognition milestone
microsoft.com
6 min read · From 2017 · By Xuedong Huang, Technical Fellow, Microsoft Last year, Microsoft’s speech and dialog research group announced a milestone in reaching human parity on the Switchboard conversational speech… · Shared by 35, including Chris Messina, Merkstatt, Dominik Grolimund, Thomas Power, Stephanie A Kowalski
techxplore.com
Physiological signals could be the key to 'emotionally intelligent' AI, scientists say
techxplore.com
3 min read · Mar 31st · Speech and language recognition technology is a rapidly developing field, which has led to the emergence of novel speech dialog systems, such as Amazon Alexa and Siri. A significant milestone in the… · Shared by 105, including Harold Sinnott 📲 #DigitalTransformation #DTW, Brian Ahier, ipfconline
TechCrunch
OpenAI open-sources Whisper, a multilingual speech recognition system
TechCrunch
2 min read · Sep 21st · OpenAI has released Whisper, a robust speech recognition model that can understand and transcribe multiple languages. · Shared by 31, including Yves Mulkers, Marcus Borba, Christophe Jean, Robert Scoble, Rob Tiffany #IoT #DigitalTwins, Chris Heilmann, Evan Kirstel the $B2B Techfluencer
Real Python
The Ultimate Guide To Speech Recognition With Python
Real Python
20+ min read · From 2018 · An in-depth tutorial on speech recognition with Python. Learn which speech recognition library gives the best results and build a full-featured "Guess The Word" game with it. · Shared by 22
awni.github.io
Speech Recognition Is Not Solved
awni.github.io
6 min read · From 2017 · Ever since Deep Learning hit the scene in speech recognition, word error rates
have fallen dramatically. But despite articles you may have read, we still
don’t have human-level speech recognition. Speech recognizers have many failure
modes. Acknowledging these and taking steps towards solving them is critical to
progress. It’s the only way to go from ASR
which works for some people, most of the time to ASR which works for all
people, all of the time.
· Shared by 28, including Ferit (at 🏠) 🌙, Rezha Julio, Thomas Power, Alexander Seifert
OpenAI
Introducing Whisper
OpenAI
1 min read · Sep 21st · We’ve trained and are open-sourcing a neural net called Whisper that approaches human level robustness and accuracy on English speech recognition. Read Paper View Code View Model Card Whisper… · Shared by 176, including Alexander Seifert, Kent Bye VoicesOfVR, Peter Steinberger, Nando 💪🇦🇺🇦🇷, Pete Warden, AGI.Eth | AIKing.Eth, Brian Roemmele, Thomas Telandro, MIT CSAIL, Danielle! 🌩, Yves Mulkers, Tactical Tech, Vikram Dutt, Sam Altman, Steve Troughton-Smith, Nico Müller 🇺🇦, Vala Afshar
Trending
OpenAI
Introducing Whisper
OpenAI
1 min read · Sep 21st · We’ve trained and are open-sourcing a neural net called Whisper that approaches human level robustness and accuracy on English speech recognition. Read Paper View Code View Model Card Whisper… · Shared by 176, including Alexander Seifert, Kent Bye VoicesOfVR, Peter Steinberger, Nando 💪🇦🇺🇦🇷, Pete Warden, AGI.Eth | AIKing.Eth, Brian Roemmele, Thomas Telandro, MIT CSAIL, Danielle! 🌩, Yves Mulkers, Tactical Tech, Vikram Dutt, Sam Altman, Steve Troughton-Smith, Nico Müller 🇺🇦, Vala Afshar
TechCrunch
OpenAI open-sources Whisper, a multilingual speech recognition system
TechCrunch
2 min read · Sep 21st · OpenAI has released Whisper, a robust speech recognition model that can understand and transcribe multiple languages. · Shared by 31, including Yves Mulkers, Marcus Borba, Christophe Jean, Robert Scoble, Rob Tiffany #IoT #DigitalTwins, Chris Heilmann, Evan Kirstel the $B2B Techfluencer
The Gradient
The Future of Speech Recognition: Where Will We Be in 2030?
The Gradient
9 min read · Aug 20th · The last two years have been some of the most exciting and highly anticipated in Automatic Speech Recognition’s (ASR’s) long and rich history, as we saw multiple enterprise-level fully neural network… · Shared by 29, including Theodora (Theo) Lau - 劉䂀曼 🌻, Dr. Marcell Vollmer 🇺🇦 #StaySafe & Carpe Diem, Nicolas Babin, Evan Kirstel the $B2B Techfluencer, Nico Müller 🇺🇦
Slashdot
OpenAI Open-Sources Whisper, a Multilingual Speech Recognition System
Slashdot
1 min read · Sep 25th · Speech recognition remains a challenging problem in AI and machine learning. In a step toward solving it, OpenAI today open-sourced Whisper, an automatic speech recognition system that the company… · Shared by 6
More like this
speechtext.ai
Domain-Specific Speech Recognition and Audio Transcription
speechtext.ai
5 min read · From 2020 · AI software for speech to text conversion and audio/video transcription. Get accurate results using domain-specific speech recognition technology! · Shared by 19, including Abel Caballero Díaz, David Garner, Jonathan Kogan
EdSurge
An Edtech User’s Glossary to Speech Recognition and AI in the Classroom
EdSurge
5 min read · 2021-09-02 · In a recent white paper, former Scholastic president of education Margery Mayer dubbed 2021 the “year of speech recognition” in education. And she may ... · Shared by 24, including Monica Burns, Ed.D. #EdTech
Google AI
An All-Neural On-Device Speech Recognizer
Google AI
4 min read · From 2019 · Posted by Johan Schalkwyk, Google Fellow, Speech Team In 2012, speech recognition research showed significant accuracy improvements with ... · Shared by 29, including Kevin Marks, Azeem Azhar, BrendanEich, Benedict Evans, Rafael Zeier, Pete Warden
VentureBeat
Google’s speech recognition technology now has a 4.9% word error rate
VentureBeat
2 min read · From 2017 · Google CEO Sundar Pichai today announced that the company’s speech recognition technology now has achieved a 4.9 percent word error rate. Put another way, Google transcribes every 20th word… · Shared by 22, including Azeem Azhar, Gerd Leonhard, Merkstatt, Thomas Power
Facebook AI
wav2vec Unsupervised: Speech recognition without supervision
Facebook AI
4 min read · 2021-05-21 · To enable speech recognition technology for many more languages spoken around the globe, Facebook AI is releasing wav2vec Unsupervised, a new method to train models with no supervision whatsoever.… · Shared by 26, including Yann LeCun, AGI.Eth | AIKing.Eth, Mike Tamir, PhD, Thomas Telandro, Alexander Seifert
Business Insider
Microsoft's AI is getting crazily good at speech recognition
Business Insider
1 min read · From 2017 · Microsoft's speech recognition efforts have hit... · Shared by 21, including Thomas Power, Thomas Telandro, Natacha Quester-Séméon Salomon, janfsauer, Howard Getson
github.com
GitHub - facebookresearch/wav2letter: Facebook AI Research Automatic Speech Recognition Toolkit
github.com
From 2018 · wav2letter - Facebook AI Research Automatic Speech Recognition Toolkit · Shared by 13, including Marco Unternaehrer, Alex J. Champandard, Soumith Chintala
code.fb.com
Wav2letter++, the fastest open source speech system, and flashlight
code.fb.com
2 min read · From 2018 · Wav2letter++ is the fastest state-of-the-art end-to-end speech recognition system available. We're also releasing flashlight, a fast, flexible ML library. · Shared by 43, including Alexander Seifert, Nico Müller 🇺🇦, Marco Unternaehrer, Elena Neira
Google Cloud
Cloud Speech-to-Text
Google Cloud
2 min read · From 2018 · Cloud Speech-to-Text provides fast and accurate speech recognition, converting audio, either from a microphone or from a file, to text in over 120 languages and variants. · Shared by 21, including Tauno, Ethan Anderson
