Connecting chemistry to pharmacology has been an objective of Guide to PHARMACOLOGY (GtoPdb) and its precursor the International Union of Basic and Clinical Pharmacology Database (IUPHAR-DB) since 2003. This has been achieved by populating our database with expert-curated relationships between documents, assays, quantitative results, chemical structures, their locations within the documents, and the protein targets in the assays (D-A-R-C-P). A wide range of challenges associated with this are described in this perspective, using illustrative examples from GtoPdb entries. Our selection process begins with judgments of pharmacological relevance and scientific quality. Even though we have a stringent focus for our small-data extraction, we note that assessing the quality of papers has become more difficult over the last 15 years. We discuss ambiguity issues with the resolution of authors' descriptions of A-R-C-P entities to standardized identifiers. We also describe developments that have made this somewhat easier over the same period both in the publication ecosystem and recent enhancements of our internal processes. This perspective concludes with a look at challenges for the future, including the wider capture of mechanistic nuances and possible impacts of text mining on automated entity extraction.