The DGS-Korpus approach to including frequent sign combinations in a corpus-based electronic sign language dictionary

With the availability of corpus data new kinds of information on sign use can be found and made available in sign language dictionary entries, such as different types of typical sign combinations (in the following called patterns) including phrasal units, loan compounds, collocational patterns and even patterns of semantic preference. As entries of SL dictionaries become more and more complex and include a wider variety of information types such as SL examples, the issue of sign representation becomes more pressing. 
For the compilation of the corpus-based DGS-German dictionary frequent sign combinations are extracted from the DGS-Korpus in order to include information on typical patterns into the entries. The dictionary displays properties of a genuine monolingual dictionary with entries including information on senses, usage, grammar, form variants, regional and age variation, synonyms and antonyms, relevant collocations and other patterns. Entries are compiled in a preliminary format and are published as work in progress step by step.
In this paper we focus on patterns of sign combinations and how to integrate them into a complex entry structure. Information types that are genuine elements or stretches of the object language have to be represented in the entry in a directly accessible way. The issue of sign representation touches the core of SL lexicography: Object language elements that fulfil the function of being relevant information in themselves (i.e. that function as examples in the broader sense) need to be present first hand in the entry and cannot be substituted.
Instead of using glosses, writing, notation or film clips, we experiment with moving icons (called micons) combined with unique entry numbers within the microstructural text of the entry. A micon is an animated image indicating the form by a still and an animation on mouse-over activation, backed up by a full-size video clip only one click away. Obviously, this is a solution solely working within the electronic medium and difficult to represent in print. 
The patterns listed in the entries are of different levels of complexity:  
1. Simple collocations of two signs, the lemma sign and a frequent neighbour, represented by two micons side by side.
2. Complex patterns where the place of the neighbour is shared by a cluster of a limited number of lexical variants, represented by a dendrogram with the micons as nodes.
3. Complex patterns where the neighbour can be any member of a grammatical class or a semantic group (colligation and semantic preference). Here the class or group can be large or even open e.g. number signs. The members of the neighbouring group are not lexical variants but differ in meaning. In our pattern representation, we include prominent members as examples as well as some categorical information on the group or class.