The semantic and stylistic differentiation of synonyms and near-synonyms

If we want to describe the action of someone who is looking out a window for an extended time, how do we choose between the words gazing, staring, and peering? What exactly is the difference between an argument, a dispute, and a row? In this paper, we describe our research in progress on the problem of lexical choice and the representations of world knowledge and of lexical structure and meaning that the task requires. In particular, we wish to deal with nuances and subtleties of denotation and connotation--shades of meaning and of style--such as those illustrated by the examples above. We are studying the task in two related contexts: machine translation, and the generation of multilingual text from a single representation of content. This work brings together several elements of our earlier research: unilingual lexical choice (Miezitis 1988); multilingual generation (R6sner and Stede 1992a,b); representing and preserving stylistic nuances in translation (DiMarco 1990; DiMarco and Hirst 1990; Mah 1991); and, more generally, analyzing and generating stylistic nuances in text (DiMarco and Hirst 1993; DiMarco et al 1992; MakutaGiluk 1991; Maknta-Giluk and DiMarco 1993; BenHassine 1992; Green 1992a,b, 1993; Hoyt forthcoming). In the present paper, we concentrate on issues in lexical representation. We describe a methodology, based on dictionary usage notes, that we are using to discover the dimensions along which similar words can be differentiated, and we discuss a two-part representation for lexical differentiation. (Our related work on lexical choice itself and its integration with other components of text generation is discussed by Stede (1993a,b, forthcoming).) aspects of their usage. 1 Such differences can include the collocational constraints of the words (e.g., groundhog and woodchuck denote the same set of animals; yet Groundhog Day, * Woodchuck Day) and the stylistic and interpersonal connotations of the words (e.g., die, pass away, snuff it; slim, skinny; police oI~icer, cop, pig). In addition, many groups of words are plesionyms (Cruse 1986)--that is, nearly synonymous; forest and woods, for example, or stared and gazed, or the German words einschrauben, festschrauben, and festziehen. ~ The notions of synonymy and plesionymy can be made more precise by means of a notion of semantic distance (such as that invoked by Hirst (1987), for example, lexical disambiguation); but this is troublesome to formalize satisfactorily. In this paper it will suffice to rely on an intuitive understanding. We consider two dimensions along which words can vary: semantic and stylistic, or, equivalently, denotative and connotative. If two words differ semantically (e.g., mist, fog), then substituting one for the other in a sentence or discourse will not necessarily preserve truth conditions; the denotations are not identical. If two words differ (solely) in stylistic features (e.g., frugal, stingy), then intersubstitution does preserve truth conditions, but the connotation--the stylistic and interpersonal effect of the sentence--is changed,s Many of the semantic distinctions between plesionyms do not lend themselves to neat, taxonomic differentiation; rather, they are fuzzy, with plesionyms often having an area of overlap. For example, the boundary between forest and wood ’tract of trees’ is vague, and there are some situations in which either word might be equally appropriate. 4