Antonymy as a collocational relation: analysis and implications for lexicographic resources

Depending on the theoretical and methodological perspective taken, antonymy can be defined as comprising a wide range of meaning opposition situations (Kagan, 1984; Miller, 1990; Justeson & Katz, 1991; Muehleisen, 1997) or, in turn, it can be used to describe a rigid connection between specific word forms (Cruse, 2000; Vossen, 2002). However, regardless of theoretical postulates and assumptions, it is consensual that collocations are helpful to find well-established antonyms and can be used to categorize them (Muehleisen, 1997; Jones, 2002; Lee, 2013). Going a step further, we advocate that an antonymy relation is truly a collocational relation, i.e. a relation defined by collocational properties: it depends on specific meaning properties (dichotomous contrast (Lyons, 1977); ‘oppositeness of meaning’ (Jackson, 1988:75)) as so many other lexical-semantic relations such as hyponymy, synonymy, etc., but it requires a high degree of ‘textual attraction’ and collocational frequency, not required or found in other relations (Amaro, 2014), fulfilling Sinclair’s 1991 idiom principle. This explains the way antonyms are processed and recognized, and why antonym pairs can vary diachronically but also geographically: we acquire specific antonym pairs by being subject to specific collocations, i.e. because these occur in specific contexts with statistic relevance in our linguistic community. This assumption can also help lexicographers deal with the theoretical and methodological challenges of accounting for antonyms, in particular in what concerns establishing the boundaries of what is useful and should be registered for a given target audience in specific lexicographic resources, e.g. monolingual, bilingual, learners’, etc. The paper will present a further look into collocation analysis and selection for antonymy extraction and treatment, departing from real examples and focusing on the use of collocational information in dictionary entries, constituting a tangible contribution to lexicographic work based on real and large data.