12.3.3 Ontological issues in thesauruses, especially in print form

Roget’s classification of English words by association with six superconcepts was designed to be exhaustive, so that each word could be found in its semantic niche within the hierarchical structure. Like other thesaurus-developers, he created his own multilayered semantic hierarchy (i.e. ontology) in the absence of any conventional model. The best-known examples of European thesauruses differ from Roget and from each other at the highest level: Hallig and Wartburg’s (1963) hierarchy for French has three supercategories: L’Univers, L’Homme, L’Homme et L’Univers, whereas Cesares’s for Spanish has just two: Dios, El Universo. However El Universo has many subcategories, for the organic and inorganic world, plants, animals, humans, individual mind/action/communication, society and its institutions (Ilson 2010: 248), making the structure very asymmetrical. These divergences in the conceptual hierarchies proposed for different European languages suggest linguistic and cultural relativities in the construction of thesauruses, which were overlooked by the universal language philosophers of the seventeenth century. The ontological construction of thesauruses for English also diverge considerably, according to whether they are designed for top-down access to the language at large, or developed ‘bottom up’ from a closed corpus or dictionary macrostructure (like the Historical Thesaurus of the Oxford English Dictionary, which generates a very large miscellany of superconcepts and over 200,000 subcategories). Either way the arbitrariness of the conceptual structure of the thesaurus is a disadvantage, making words less easy to find than Roget intended. In fact John Roget added an alphabetical index at the back of the family Thesaurus in 1879 (ten years after his father’s death), and it is now a standard feature with onomasiological thesauruses. Roget’s Thesaurus and those modelled on it are notionally synchronic representations of the English lexicon (Ilson 2010: 251). Yet readers at the turn of the twenty-fi rst century would be aware that some of the words contained in Roget’s sets of alternatives are rare or obsolescent, e.g. adumbration, alack, lachrymose, sedulity, tergiversation. This problem is addressed in the Historical Thesaurus of the Oxford English Dictionary (HTOED) published in 2009, which gives the fi rst recorded date of each near-synonym and notes those which are obsolete. Drawing on the microstructural resources of the OED, HTOED is also able to provide stylistic information about the usage of words, e.g. booze as a colloquial term for ‘wine’, and plonk as an Australian/British word for ‘cheap wine’, though both are listed in a standard thesaurus without such labels, or definitions to help discriminate them. The onomasiological construction also masks the natural polysemy of common words, since the different senses of the same word have to be included in different subcategories, e.g. horse in both ‘animal’ and ‘framework’. There is no cross-referencing, so again it is the index rather than the ontological structure which provides this information. Both synchronic and diachronic thesauruses are paradigmatic rather than syntagmatic (Ilson 2010: 256) in their projection of the lexicon. Neither shows the collocational tendencies that contribute to the meanings of words and help to discriminate their senses,

e.g. the different senses of the word freedom to be found in freedom from and freedom to. Other scholars have noted that thesauruses do not indicate the transitivity relations that verbs enter into, which help to differentiate transitive and intransitive/mediopassive uses (Burger 1991: 49–52). Compare ‘he sells books’ with ‘the book sells well’. The relative frequencies of words, now easily extracted from reference corpora and reported in learners dictionaries (e.g. Collins Cobuild and LDOCE), are not indicated in either the synchronic or diachronic thesaurus. While modern print dictionaries accommodate all these aspects of lexical meaning within the microstructure of individual entries, it is still unclear how/where they could be incorporated into the macrostructure of the print thesaurus.