You're always (always!) dealing with many (many!) tables - with Madelon Hulsebos

[@magdalenaanopsy9668]

What a great episode! Kudos to Madelon, very inspiring person.

[@EkShunya]

great !!!! lets gooooooo

[@trinityblood5622]

I have been following her research recently. As I am doing data discovery and integration research. Her PhD Thesis is very well written. But the problem with Tabular RL is that traditional embedding models/architecture for words/sentences will not do good for tabular data.

[@madelonhulsebos2082]

Great to hear @trinityblood5622, and that is a great point that we have have studied in several research projects (e.g. in chap 2&5 as summarized in sec 6.2.3 in the thesis). Figuring out the exact limitations of existing language models and developing better specialized architectures, as well as training data and tasks, has a high priority on the research agenda!