This paper considers the process of organising semantic information in bilingual dictionaries with diachronic coverage, from selecting the textual source-material to designing the entries. The discussion centres on practical aspects of ancient Greek lexicography. First, the traditional semantic frameworks are described. Then, more recent approaches are noted, notably those of Adrados and of Chadwick, both of which aim to integrate contextual data within a semantic framework. Since the relevance of contextual information varies with lemma part of speech, different configurations are required for entries describing nouns, adjectives, and verbs. These are illustrated by three entries from a Greek-English dictionary currently being written at Cambridge. In order to organise data to this level of specificity, stylistic templates are indispensable, and digital software provides a means of providing them. However, systems designed for writing new dictionaries require different features from those designed for encoding pre-existing texts. A description is given of how the lexicographic requirements of the Cambridge dictionary were met by a user-designed system.