Paper short abstract:

This paper analyzes the process of producing digital editions of historic dictionaries as tools for knowledge discovery based on an interplay of production technologies (XML, Text Encoding Initiative), editorial practices (representation, annotation, metadata) and semantic taxonomies.

Paper long abstract:

This paper explores the theoretical and practical concerns of producing a digital edition of Vuk Stefanović Karadžić's ethnographically rich *Lexicon Serbico-Germanico-Latinum* (1818, 1852). In addition to marking up existing elements of a dictionary entry (such as lemma, part of speech, senses, definitions, translation equivalents, examples etc), our work supplies additional information to enhance the user's interaction with the dictionary, including: a) an extension of the extant cross-reference system through linking synonymous and near-synonymous entries that have been overlooked by previous editors (e.g. жаба and напнигуша; обрљуга and неопера); b) marking up persons, places and dates for easy indexing and analysis; c) standardization of word usage labels (eg. ист. and ист. кр. as <usage type="geo">East</usg> for better statistical analysis and possible further processing and creation of geo-spatial word maps etc.); and d) introduction of semantic domain labels in accordance with Magnini and Cavaglia, 2000; Bentivogli et al., 2004. The domain labels provide valuable information about the clusters of paradigmatically related terms, as well as the extent of domain ambiguity and domain variability. Users will be able to treat semantic domains as thematic entry points into the dictionary, looking up, for instance, all entries that belong to AGRICULTURE, HISTORY or GASTRONOMY; while our logically and semantically consistent markup of Karadžić's own usage notes will make it possible to easily explore regional and dialectological distribution of entries in the *Lexicon*, offering a basis for subsequent work that could involve data visualization, statistical analysis, text mining etc.