Understanding users and their tasks and data is critical for developing successful data systems. Semantic annotation of relational tables is one of the most versatile problems in data management in this context. Effectively solving it will help with diverse tasks such as data cleaning, schema matching, semantic search, data discovery, data and visualization recommendation, automated machine learning, and so on. In this talk, I’ll give an overview of our work on semantic type detection, including models and datasets we’ve developed and curated over the years. I’ll also share the open problems and future directions for automated semantic annotation and data systems with built-in semantics support.