Representation and method space

Method space is a metaphor for the degrees of freedom available to the scientist in developing a theory for a given domain with specific methods. A lexicon, its contents, and their representation are dependent on the same three kinds of criterion:

1. Domain, divided into the two orthogonal dimensions represented by the ILEX model:
   1. Composition of lexical units in terms of ranks and constituents at each rank, from the smallest, morphemes, through stems, derived words, compound words, phrases, sentences to lexicalised discourse units;
   2. Interpretation of each of these ranks: surface interpretation involving phonology and orthography, and semantic (and pragmatic) interpretation.
2. Empirical method, defined in terms of the lexicographer's intuition and statistical investigation of the use of lexical items in large text corpora.
3. Formal method, defined in terms of theories, models, formalisms, notations, computer implementations.

This three-dimensional method space, visualised in Figure 2, provides a useful heuristic checklist for describing different approaches.

 
Figure 2: Method space for building lexica. 

The domains have already been characterised with reference to linguistic signs; the formal background to representations requires more discussion, and in the second part of this section, lexical representation using the DATR lexical knowledge representation language will be discussed from the point of view of relating lexicon theory with practical computational lexicography. Empirical methods are touched on in the section on practical lexicography.