🎉 New Year, New Innovations: Papers, Open-Source Tools, and CHiME-8 Success!
We’re kicking off 2025 with updates on latest breakthroughs in speech recognition and diarization from the BUT Speech@FIT team in collaboration with @jhuclsp
Read more in this thread 👇
[1/n]
🚨 Introducing BenCzechMark (BCM) 🇨🇿—the 1st multitask & multimetric Czech benchmark for large language models! 🧠
🔗 Check out the leaderboard: https://t.co/3nFANXN35i
📖 Read more in our Hugging Face blog: https://t.co/wxnD7BMoKn
#NLP#AI#CzechLanguage#LLM
JSALT23 team united together to make FST great once again!
The result of 2 week workshop is an open-source library with support for differentiable FST operations (e.g. shortestdistance, composition) under user-defined semiring.
https://t.co/ycxHRJIkNE
The goal is to build a new FST library with the aim on performance. The library is in Julia and FSTs are represented as sparse 4D tensors, which allows to write operations like shortest distance or composition in terms of linear algebra operations.
[2/3]
Exciting news! 🎉 #SpeechBrain 1.0 is out with tons of thrilling advancements.
Our #OpenSource toolkit now features 200+ recipes and 100+ pretrained models on #HuggingFace for diverse #ConversationalAI tasks.
🌐 Website: https://t.co/a1wqxLucgw
💻 Repo: https://t.co/MsCZbSbSOf
If you watch this space, you already know my love for the neural transducer. I skimmed through all 21 papers relating to transducers that were presented at #INTERSPEECH2023, and wrote a summary blog: https://t.co/nBXivyFjml
Summary in 5 bullets:
🚨 ring ring ring 🔔
Our work on Accent Classification with #CommonVoice dataset was accepted at @ISCAInterspeech 2023!
"CommonAccent: Exploring Large Acoustic Pretrained Models for Accent Classification Based on Common Voice"
abs: https://t.co/9AJ0lxwXm9
[1/N]
10 years ago, WFST-based methods were the norm for speech processing (think, Kaldi).
Since then, end-to-end models have become quite the rage --- they are simple, do not require much domain expertise, and you can train a PyTorch model for a new task over a weekend. ⚡️
1/n
Hello! Tomorrow is this year’s JSALT last day. In the afternoon, starting 2pm CET, we will be discussing FSTs and lattices for speech recognition and downstream tasks along with @rdesh26, @Umberto_Senpai, @MartinKocour and a big lot of other people https://t.co/XXpmdNocTt
Are you missing WFSTs in ASR? Today we are presenting modern WFST toolkit called TensorFSTs.jl, where all WFST operations are based on Linear Algebra and are differentiable.
Our #JSALT23 presentation is planned on 2pm(CET) and will be streamed on YouTube (link below). Join us!
BUT has a strong presence at JSALT2023 Jelinek Summer Workshop on Speech and Language Technology. The final presentations are now streamed on youtube (link below).
The event is co-sponsored by the European project #ESPERANTO coordinated by @LeMansUniv.
https://t.co/bvgZA1c3c5
Pleasure to connect with like-minded speech recognition scientists!
Special shoutout to everyone who showed interest into our WFST exercise in Julia. Your enthusiasm and engagement made it even more rewarding!
📓Lecture + Exercise:
https://t.co/n8TlkLToiG
The JSALT summer school wraps up tomorrow. On Monday, the workshop starts. Looking forward to 6 weeks of WFSTs + speech!
Pic from: https://t.co/a1AZYJSdU2
@Pablogomez3 why scared? there is nothing to be scared of... OpenAI marketing just know how to hide their know-how. The name of the company is also a joke.
To be clear: I'm not criticizing OpenAI's work nor their claims.
I'm trying to correct a *perception* by the public & the media who see chatGPT as this incredibly new, innovative, & unique technological breakthrough that is far ahead of everyone else.
It's just not.
#SpeechBrain reached 5k stars on #Github!
This is an important milestone for our community. Let me thank all the contributors!
The team is working hard for the next major release.
https://t.co/a1wqxLuK64