The paper offers a discussion of the principles and practicalities behind the Bulgarian Sense-Annotated Corpus (BulSemCor), presents the results, and sketches the challenges encountered in the process of annotation, the adopted conventions and the decisions made. First, the corpus structure and the tool for annotation are presented in brief, followed by a discussion of the methodology for identification and annotation of different types of language units, the strategies towards challenging phenomena with respect to part-of-speech and morpho-syntactic classification, the approaches for handling certain syntactic phenomena such as elliptic constructions and coordinate compound words, etc. The encoding of language-specific concepts and the decisions with respect to the organisation of BulNet (the lexical-semantic net that provides the inventory of senses for annotation), are also covered. Finally, the corpus applications and future developments are outlined.