Automatic Precisiation of Meaning

Semantic similarity measures play a crucial role in providing a level of understanding of word semantics, essential for adaptive and robust interactions in phenotropic interfaces. However, semantic similarity only represent the meaning of a word related to another, lacking thus a general understanding of the semantics of the word itself. To address this limitation and achieve a broader comprehension of word meaning in language, the computing with words (CWW) pipeline is employed. This offers a practical framework for interpreting and treating human perceptions in a machine-understandable way. The precisiation phase of the CWW pipeline involves translating the semantics of words into their mathematical representation by means of fuzzy sets, on which computations can be performed to retrieve meaningful answers. However, the existing methodology for precisiation necessitates manual interventions and crowdsourcing. This hinders the flexibility, adaptivity, and robustness of any phenotropic interface building on CWW. To overcome these limitations, this chapter presents two iterations of an automated method for precisiating the meaning of scalar adjectives and adverbs. Building on a spectral semantic similarity measure, this approach reduces the reliance on manual interventions for the precisiation of meaning. An evaluation of the proposed algorithms proves their efficacy at performing precisiation of meaning with an accuracy similar to that obtained by means of manual precisiation.