One of the focuses of this year’s @trlworkshop is on insights from industry 👀. Is your startup/team/org interested in AI for structured data? Consider sponsoring the Table Representation Learning workshop to support this emerging community. DM/email @madelonhulsebos by 30 Aug✨

Looking forward to VLDB 2024 next week! 🇨🇳 I will be presenting our work on data quality assertions for LLM pipelines. At least 5 of us Berkeley students & postdocs will be there! I will also be visiting HKUST next Friday & would love to meet folks in data, HCI, and/or ML.

🔍 Unsure which table embedding model to use? Super excited to introduce our work Observatory: Characterizing Embeddings of Relational Tables, which will be presented by myself in VLDB 2024 this Thursday. Feel free to stop by Session C7 13:30-15:00 or DM me if you are interested

🌟 Key contribution: We introduce a set of properties motivated by the relational data model (e.g., structural invariance) and practical factors to consider when using embeddings (e.g., sample size and context changes) to analyze embedding representations of relational tables.

🍇 Fruitful collaboration with Madelon Hulsebos, Zhenjie Sun, Paul Groth, and my adorable advisor Jag Paper: arxiv.org/abs/2310.07736 Codebase: github.com/superctj/obser… Python library for easy embedding inference: github.com/superctj/obser…

The first submissions are in 👀… don’t miss out on this exciting edition of @trlworkshop at NeurIPS 2024! Submission deadline is in 2 wks: 20 Sep ⌛️

Check out NUDGE, a simple and lightweight fine-tuning method for pre-trained embeddings that boosts text/image retrieval accuracy by more than 10% in minutes on a CPU! blog: data-people-group.github.io/blogs/2024/09/… paper: arxiv.org/abs/2409.02343 code: github.com/szeighami/nudg…

Vincent D. Warmerdam I’m actually listening to your Madelon episode rn. Really relevant and great guest!

New podcast episode is out! This one is all about the fact that in real life, you are always dealing with many (!) tables. So maybe we need better tools to find the right ones? We talk to Madelon Hulsebos about it, who has a Phd on the topic! eu1.hubs.ly/H0cgsVc0

Really nice conversation w Vincent about (multi-)table retrieval (eg for dataset search, rag), and how (table) representation learning & generative models can help in surfacing insights from structured data end-to-end, efficient and reliable. Thanks for the chat Vincent D. Warmerdam!!

Is it finally time to rewrite and -print the US Constitution for when ..she.. takes office? 😎