A sintaxe-semântica dos adjetivos predicativos no português brasileiro: uma proposta taxonômica

Abstract

Syntactic, semantic and lexical classifications are useful resources for research, computational processing and teaching of Portuguese. They include several syntactic and often semantic characteristics of various lexical items, e.g., whether they accept human or clausal subjects, complements and so on. For linguistics as a field, they are particularly useful to establish relations between the structures available for each predicate and their lexical semantics. It is thus possible to verify whether items with similar distributions also present similar meaning. Items such as claro (clear), evidente (evident), óbvio (obvious) and curioso (curious) accept clausal subjects (“É claro/evidente/óbvio/curioso que o João fez isso”, “It is clear/evident/obvious/curious that João did that”), while items such as redondo (round), cheiroso (nice-smelling) and vermelho (red) do not (“*É redondo/cheiroso/vermelho que o João fez isso”, “It is round/nice-smelling/red that João did that”). Syntactic or lexical-semantic and syntactic classifications of predicative adjectives have been elaborated for other languages, such as French, Italian, Korean and Malagasy, in addition to European Portuguese, but not for Brazilian Portuguese. This work presents a descriptive linguistic study aimed at providing a taxonomy of Brazilian Portuguese adjectives based on their syntactic and lexical-semantic properties. To do so, it resorts to Lexicon-Grammar, a theoretical and methodological framework for linguistic description based on Zellig Harris’ Transformational Operator Grammar. This approach postulates that sentences be described as relations between operators and their arguments; these undergo several types of transformations. We extracted seven thousand lemmas automatically identified as adjectives in a corpus of written texts from the internet, comprising news articles, institutional content, and user-generated texts, compiled in 2020. These lemmas were revised and analyzed through intuition and corpora consultation, leading to 7,106 adjective entries, 3,161 of which were considered predicative adjectives that cannot be regularly derived from verbal constructions. For each of these adjectives, 36 syntacitc properties were registered. This led to a classification based on the distributional properties of their arguments. We proceeded then to analyze items having formal features in common, considering their lexical-syntactic similarity. Some patterns common to certain formal classes of adjectives were observed, but they present several exceptions and irregularities. We conclude that the lexical semantics of Brazilian Portuguese adjectives have no necessary and sufficient relations with formal properties; however, several items with similar formal features also share similar senses.