Benedetta Cevoli retweeted
Accuracy is THE definitive benchmark of quality for speech technology 🎯
Without it, you lose trust in your transcription, and any process built with it might be susceptible to compounding errors.
More errors = less trust
Fewer errors = more trust, more quality, more value.
But don't just take our word for it.
@ArtificialAnlys, an independent third party conducting benchmarks on AI systems, has independently assessed all leading speech vendors.
And their latest results reveal....
....
....
Speechmatics is the definitive market leader on accuracy, with the lowest WER available🥇🥇🥇
When close enough isn't good enough - choose @Speeechmatics
Check out the full breakdown 👇🔗
artificialanalysis.ai/speech…
Congratulations to @Speechmatics on setting a new record in our Word Error Rate benchmark for Speech to Text models!
Speechmatics’ Enhanced model has achieved a Word Error Rate of 6.5%, the best result yet among the models we test. Accuracy is most important for use-cases involving audio that might be noisy, of variable quality and involving a wide range of speakers.
For full details, see our Speech to Text leaderboard: artificialanalysis.ai/speech…
Benedetta Cevoli retweeted
This is the next frontier of AI 🦾
Bringing together the best of Speech Recognition and breakthroughs in LLMs, voice is now the biggest untapped resource in the world.
This is foundational speech technology for the AI era
👉 bit.ly/3SlXCFt
Benedetta Cevoli retweeted
🎙️📷 Interspeech Day 1!
This year, Speechmatics is providing captions at #Interspeech2023, boosting accessibility for all attendees. Meet the team behind our AI-driven speech recognition.
Book a meeting here to connect at the conference bit.ly/3shDjOG 🗓️🚀
Benedetta Cevoli retweeted
The @Speechmatics crew is at #interspeech2023 !
Benedetta Cevoli retweeted
If you don't have time to watch the full video of @sama in congress, I used @Speechmatics to transcribe the full video and the latest @AnthropicAI Claude 100k Context Window to summerise the whole transcript:
Benedetta Cevoli retweeted
Speechlings, @BCevoli and Bethan Thomas hosting the @SHaPS_UCL Speech Science Forum today 🙌🗣️🗯️
Benedetta Cevoli retweeted
Repeated this test with new Speechmatics. WER 6.88. Indeed new Ursa model improved a lot!
Benedetta Cevoli retweeted
We are proud to share that Speechmatics has been named No.5 in @FastCompany’s 10 Most Innovative Companies in Artificial Intelligence of 2023 with the likes of @OpenAI, @DeepMind, and @Builderai 🤯🥳
#FCMostInnovative here: bit.ly/41zfM9f
Benedetta Cevoli retweeted
There’s a big difference between speech-to-text working for some people, most of the time, and for all people, all of the time 🔇
@BCevoli looks at why speech tech in 2023 still has some distance to travel in reducing inequality ➡️ bit.ly/3i54xU7
Benedetta Cevoli retweeted
🎉Congratulations to everyone involved in the @KU_Leuven Summer School on the #Law, #Ethics & #Policy of #AI. 9 days of an intense programme have come to an end with a lot of discussions had, insights gained, and friends made. Looking forward to the next edition.
That’s a wrap!
What a privilege and pleasure it was to host such a wonderful & talented group of people in Leuven for the @KUL_AI_SSchool, both virtually & physically😊
A massive thank you to all the participants & lecturers, and to @GAVClaw & @Hendrickx_Vic! See you all soon!
Benedetta Cevoli retweeted
A wonderful overview on AI & the public sector by @NathalieSmuha at @KUL_AI_SSchool, discussing applications, values and pitfalls.
“Digitizing a broken process gets you a digitized broken process”. Thank you for being so good Nathalie!
(Yes, this school is never ending 😂)
Benedetta Cevoli retweeted
That’s a wrap!
What a privilege and pleasure it was to host such a wonderful & talented group of people in Leuven for the @KUL_AI_SSchool, both virtually & physically😊
A massive thank you to all the participants & lecturers, and to @GAVClaw & @Hendrickx_Vic! See you all soon!
Benedetta Cevoli retweeted
**New paper** on use of prediction in skilled text reading with @bcevoli and @snobfox. Main finding: accurate predictions can be made based on text statistics, and these predictions guide how our eyes interrogate upcoming text. THREAD 1/ #GPT2 royalsocietypublishing.org/d…
Benedetta Cevoli retweeted
Featured in this month’s @TechnologyMagBC 💥
“AI should be honest, trustworthy at all levels - from its inception to deployment” - @BCevoli
Check out the article 📖 (pages 128 - 135) bit.ly/3bPJVw1
Benedetta Cevoli retweeted
At the height of the pandemic, eLearning tools accelerated the use of #transcription software. Can we, however, rely on it to educate the next generation if it misunderstands young voices? 👶 bit.ly/3rwGkqj
Benedetta Cevoli retweeted
🔹 What does the term "bias" in #AI mean?
🔹 How can we make a difference?
🔹 How accuracy can lead the way
Explore the answers on the blog by our very own @BCevoli ✍️ bit.ly/3DPc1iS
Benedetta Cevoli retweeted
** New Preprint ** where we show that internal states of GPT-2 predict fine-grained eye-movement behaviours in reading: skips, fixation durations, regressions. Excited about this work with @BCevoli and Chris Watkins psyarxiv.com/urjh4
Benedetta Cevoli retweeted
When technology is at its best, every voice should be treated equally. How does this work in reality? Learn more about the research findings that prompted us to act via our whitepaper 📄 bit.ly/2ZciOVp #aibias
Benedetta Cevoli retweeted
Our research into #AIbias and inclusion in speech recognition propelled us into action. Read more about the results that fueled our latest technology: bit.ly/2ZciOVp
Benedetta Cevoli retweeted
Yesterday we announced a major #AI breakthrough and how we are tackling #AI bias, inclusivity, and diversity in speech recognition. Find out more via @TechCrunch: tcrn.ch/3pHxjvm





























