February 2023 Community News: New audio labeling tutorial, Academic Program, and PyCon!

Guide to Labeling Audio Data

Follow along with the written tutorial and video to label audio data, and try the latest features in the Label Studio 1.7 release, including enhanced sound detection, new selection methods, advanced timeline zoom (up to x1500 times), and more fine-tuned controls.

Apply for the Academic Program

Are you using Label Studio for academic research? See if you qualify for free access to Label Studio Enterprise as a cloud service as part of our new academic program.

Improving on RLHF with Language Feedback

Chris Hoge explores the basics of RLHF, how it was used in building models like ChatGPT, and the direction new research is taking us in human-guided training and refinement.

Upcoming Events

PyData
April 17-19
Berlin, Germany

Come meet CTO Nikolai Liubimov and Sr. Technical Community Advocate Erin Mikail Staples, who will be speaking about RLHF.

PyCon
April 21-23
Salt Lake City, UT

Are you attending PyCon? Let’s meet up! Let us know in the community slack #08-ls-events, or get in touch with Chris Hoge.

Community Shoutouts

Kudos to Vadim Shestopalov, Greta Attard, and Maxine Lyu for being outstanding community contributors! They’ve all helped out with both upstream contributions and reporting bugs, but also helping out other community members.

We love to see folks like JP de Vooght and Marc Ivan Manalac building in public as they use Label Studio.

As a reminder, we’ve launched a pilot program for content creators, and we’d love for you to contribute!

Annotations

Oscar season is coming up, and to celebrate, we share this Information is Beautiful article that explores what the most successful movie of all time might be. If you measure by gross revenue, even adjusted for inflation, it’s Avatar. However, if you dig deeper into the data and consider other metrics of success, the results might surprise… and terrify you.

RLHF has been on our mind, and apparently, everyone else’s too. If you want to learn more about this important training technique, Hugging Face has a great introduction to how it works. Meanwhile, The Sequence published an explainer on how RLHF was used to train ChatGPT.

The team at DataJournalism.com just released the results of their second annual State of Data Journalism, along with thefull data from the survey. If you’re interested in the ever-expanding landscape of data journalism and emerging trends, this is an excellent resource to check out.