I’m looking for feedback and interest on a 1 or 2 week crash course in programming and AI / machine learning aimed at documentary linguists. Would you attend? What factors would affect your decision? The idea is to make it free but all travel and lodging costs fall to attendees. And if it has good attendance and good feedback, then to propose it for an event like COLANG. Below is the description and under that is a link to the draft syllabus.
This two-week workshop will teach concepts and skills underlying AI. The primary objective is that participants will be enabled to lead teams or collaborate closely with technologists for the purpose of building technology for languages with limited resources. Lectures and exercises cover basic Python programming and the general mathematical principles underlying machine learning which is the main engine of modern AI. Participants will apply what they learn by collaboratively improving a machine learning model trained on limited language data. The intended audience is native speakers, linguists, and anyone interested in natural language processing (NLP) or AI for endangered or low-resource languages.
Even if I can’t attend in person as a volunteer “tech coach”, would be happy to help out in some of the groundwork like the start script/Colab for ASR fine-tuning.
Based on survey of confirmed attendees (especially those who can’t work on their own data on Colab), also happy to help prep non-English test and perhaps train data from a handful of DoReCo languages
Could the materials be made openly available after the workshop both as notebooks/scripts and perhaps even recordings for upload to YouTube (e.g. Stanford 224N: 2023, 2021, 2019). Is that already the idea behind BELT?
BELT provides straightforward tutorials that teach skills for building real, practical language tools for your own language. You don’t need a programming or linguistics background, just some text from your language and knowledge about it.
Thank you @fauxneticien! Good thoughts. Having help with ASR and general CoLab scripting would be highly appreciated! And I think short online sessions with a “tech coach” is great idea (all day online is too much, which is why I’m proposing an in-person event)
The material for both weeks is already publicly available, except for the last day on LLMs which I have not developed yet. You should be able to follow the links from the Google Doc. The second week material is not well prepared. Someday I’d love to make it a user-friendly and self-paced curriculum, like BELT.
I read the Google Docs document and thought I should comment on the intended audience. The use of “native speaker” is unfortunate as there are many learners who work with mother-tongue and other extremely proficient speakers and with leagacy recordings who could very much benefit from the proposed course. In additiona, it would be important to include language pedagogists/androgogists (both Indigneous and non-indigenous) in the mix as well. I have seen some apps created that are not much better than using text and at times worse than just watching uneditted video recordings. I have recently seen an AI “talking head” “speaking” an Indigeous language in a language learning program. It was nothing more than a gimic as the AI generated avatar was not showing how to articulate the words properly. It was like a poor audio dubbing but with AI created video.