From Surface to Semantics: Presenting DoTabler at ECAI 2025

Last October, I had the privilege of presenting our paper, From Surface to Semantics: Semantic Structure Parsing for Table-Centric Document Analysis, at the 28th European Conference on Artificial Intelligence (ECAI 2025) in Bologna, Italy. Standing in front of researchers whose work I had only ever read about in citations was both surreal and humbling.

The three days flew by. ECAI brings together work from across the AI spectrum — from formal logic and reinforcement learning to natural language processing and explainable AI — and the diversity of perspectives was exactly what I had been hungry for as a junior PhD. I sat in on talks far outside my area, exchanged ideas over coffee with people whose problems were nothing like mine, and walked back to my hotel each evening with my notebook fuller than when I left it that morning.

The questions after my talk were sharp but generous. A few left me thinking about ideas in our work from angles I hadn’t considered, and one corridor conversation has already turned into a follow-up project I’m excited to share later this year.

Beyond the science, ECAI was the kind of trip that reminds you research is also a community. The conference social dinner brought together scholars from across continents around long shared tables — and across plates of tagliatelle al ragù, mortadella, and other Bologna classics, conversations drifted between research, cities, and the small detours that bring people to the same room. I came home with a stack of business cards, a longer list of papers to read than ideas of my own, and the quiet reminder, gratefully accepted, of how much there is still to learn.

Thank you to my coauthors and supervisors, to the ECAI organizers, and to everyone whose questions, encouragement, and casual hallway chat shaped this trip into what it was.