Lexical relatedness and the lexical entry – a formal unification

Andrew Spencer, University of Essex, UK Paradigm-based (realizational-inferential) models of morphology are defined over a notion of ‘lexeme’. Inflectional morphology defines the set of word forms which realize various morphosyntactic properties sets (MSPSs) for a given lexeme (e.g. from PRINT, {print, prints, printing, printed}. By contrast, a derivational process, say, Subject Nominalization (SubjNom), yields a new lexeme PRINTER with its own inflected word forms {printer, printers}. These are not word forms of the lexeme PRINT. In the classical PFM model Stump (2001) inflection is defined by a paradigm function (PF) which specifies the word forms realizing a given MSPS for that lexeme. Stump treats fully regular (paradigmatic) derivation in the same way, for instance, PrivativeAdjective formation is governed by a derivational feature δ which delivers the -less suffixed form: PF() = . But this approach leaves several questions open. Q1: How do we ensure that the derived lexeme inflects in the appropriate way (i.e. ensuring that PRINTER inflects as a noun rather than a verb)? Q2: How do we reconfigure the classical model to reflect the fact that derivation involves the addition of a semantic predicate, not just the realization of a morpholexical feature? Q3: How do we capture types of lexical relatedness intermediate between inflection/derivation? For instance, in many languages there are transpositions such as deverbal participles, which have the external syntax/morphology of an adjective but yet are forms of a verb lexeme, and there are several other types of intermediate case. Q4: How dowe account for the fact that derived lexemes often undergo semantic drift while still exhibiting the morphological idiosyncrasies of the regular derivate, e.g. transmit ∼ transmission (of a car)? I propose amodel of lexical representation/relatedness for use in any realization-inferential model which permits us to treat inflection, all the various kinds of derivation, and all intermediate types of lexical relatedness as the result of the same formal class of operations at a certain level of abstraction, and which defines the basic lexical entry with the same machinery (effectively as a representation which is related to itself). I assume that a lexical representation is at least a four-dimensional object as in (1)1: (1) 