Lexicon models and lexical representation

 

A given system lexicon or lexical database  is based on a lexical information model  or a data model ; often the model is intuitively constructed, or based on notions taken from traditional school grammar,   but scientifically motivated models are becoming available. A model of lexical information will make at least the following distinctions:

• Lexical objects: The basic objects (such as words) described in a lexicon. It is becoming customary in lexicography and computational linguistics to refer to the lexical sign, i.e. an object associated with attributes denoting orthogonal kinds of lexical information. A second kind of lexical object is the lexical sign class or archi-sign in which similar lexical objects are grouped together, each characterised by subsets of the lexical information required to characterise specific lexical signs. These class-based generalisations may be organised in terms of implication rules (redundancy rules),   , subsumption lattices, type hierarchies , or default inheritance hierarchies .
• Lexical information: In a theoretically well-founded lexicon which satisfies formal criteria of consistency and coverage criteria such as empirical completeness and soundness, types of lexical information are orthogonal, i.e. of different types which complement each other. These orthogonal types of lexical information are often labelled with attribute names, and the items of information regarded as the values of these attributes. Values may be complex, expressed as nested attribute-value structures. The types include orthography, pronunciation, syntactic distributional properties, meaning, and pragmatic  properties of use in context (e.g. speech act type, stylistic level). See also the results of the EAGLES Working Group on Formalisms.

Lexicon models  for lexical databases  and system lexicons are part of the overall conceptual framework required for lexicon development. Modern approaches to lexicon development provide suitable lexical representation languages for formulating and integrating the different kinds of lexical information specified in a lexicon model and assigning them to lexical objects, and implementations for these representation languages [Andry et al. (1992)]. In recent work, the following useful distinctions are sometimes made:

• Lexicon formalism: A specially designed logic programming language such as DATR, or an algebraic formalism such as attribute-value matrices, or appropriate definitions in high level languages such as LISP or Prolog, with compiler concepts for translating these languages into conventional languages for efficient processing. Imperative languages such as C are sometimes used directly to represent smaller lexicons, or where speed of access is at a premium, but this is not a generally recommended practice.
• Lexicon theory:  A coherent and consistent set of expressions formulated in a well-defined formalism and interpreted with respect to a lexicon model.
  • General lexicon theory:  A general theory of lexical objects and information, for instance a theory of lexical signs and their representation.
  • Specific lexicon theory: A given lexicon formulated in a lexicon formalism on the basis of a lexicon model.
• Lexicon model:  Specification of the domain denoted by a lexicon theory, conceptually independent of the theory itself (cf. the notion of a data model  for a database). A different definition is also common: the general structure of the objects and attribute-value structures in a formal lexicon. A lexicon model specifies the following kinds of information:
  • Types of lexical object and structure of lexical entries.
  • Types of lexical information associated with lexical objects in lexical entries.
  • Relations between lexical objects and structure of the lexicon as a whole lexicon architecture. 
• Linguistic framework: In recent large projects such as VERBMOBIL,  general linguistic frameworks such as HPSG (Head-Driven Phrase Structure Grammar)  have been used.

The aspects of representation and architecture will be dealt with in a later section. The following subsections are concerned with the main kinds of lexical information required for spoken language lexical entries.