In the declarative approach to problem solving, a widely recognised challenge is how best to capture the relevant domain knowledge in a formal knowledge base. A lot of research focuses on formalising the knowledge of domain experts, but in addition, a suitable knowledge base typically also needs to incorporate data coming from existing sources (e.g. a database or CSV file). This data is often extracted by means of ad hoc scripts, e.g. in a general-purpose imperative programming language. In this paper, we study this task from a logical perspective. We analyse how to derive a logical vocabulary from tabular data, in order to transform the data into a logical structure. As data from multiple sources often needs to be combined, this paper will further discuss the combination of multiple logical specifications. We conduct our study in the context of the IDP-Z3 reasoning engine for the FO(.) language, a rich extension of classical first-order logic. For this engine, we implement a new API called KeBAP (Knowledge-Base API for Python), which automates tasks such as deriving logical vocabulary from data tables and merging different data sources.