Under the recently established ELLIS unit Amsterdam’s research theme on Table Representation Learning, Madelon Hulsebos (CWI), Iacer Calixto (Amsterdam UMC), Michael Cochez (VU), and Andrew Yates (UvA) organized a workshop focusing on representation learning and generative models for structured data, such as relational tables and spreadsheets, as well as knowledge graphs. The workshop was held on February 27th 2025 at the CWI, Amsterdam Science Park.
The workshop hosted five speakers:
– Paolo Papotti (Eurecom): SQL and Large Language Models: A Marriage Made in Heaven?
– Effy Xue Li (University of Amsterdam): Efficient Use of LLMs for Data Preparation
– Frank Hutter (University of Freiburg / ELLIS Institute Tübingen): Accurate predictions on small data (and time series) with the tabular foundation model TabPFN
– Julian Eisenschlos (Google DeepMind): Visual language: how generation can drive understanding in data visualizations
– Pasquale Minervini (University of Edinburgh): Integrating Algorithms and Neural Models
The workshop brought in different views from the NLP, ML, DB, and IR disciplines and hosted 17 extended abstracts (https://openreview.net/group?id=ELLIS.eu/2024/Workshop/RLGMSD), on, for example, retrieval augmented generation (RAG) and question answering (QA) over structured data. The distribution of affiliation countries who submitted abstracts (not necessarily representative for the attendees) include the Netherlands (33.3%), France (19%), Germany (14.3%), Italy (9.5%), Belgium (9.5%), the UK (4.8%), and others. The workshop received positive responses with a total of 70 participants, among which are PhD candidates, master students, postdocs, and senior researchers from industry and academia.
The generous sponsorship from SAP, and the ELLIS unit Amsterdam via the Workshop Support Programme, also made it possible for the workshop to support travel of 3 master students from foreign European institutes.
