O tratamento da relação semântica partitiva em recurso lexical dedicado à legibilidade por máquina: estendendo a anotação da FrameNet

Abstract

In the scope of Cognitive Linguistics, the present work sought the key elements for the understanding of the phenomenon of signification and the proposition of formalization in an ontological conceptual model, which subsidized processes of inference in the machine interpretation of the partitive relation. The main objective was to study the partitive semantic relation as manifested in the semantics of human languages and, for this purpose, the study advances from the classical lexical perspective, limited to the expressions “is part of” or “has part” from created sentences, as “the handle is part of the cup” or “cup has the handle as part”, for real occurrences extracted from corpus, as in the sentence: “a asa da xícara quebrou” [the cup handle broke]. In the Frame Semantics approach, in order to understand a linguistic item, it is necessary to understand the frame that it evokes and the relations of this frame with other frames. In addition to the description of the partitive relation in the scope of Frame Semantics, the objective was to formalize the partitive relation between frames in an ontological model, presenting the proposal of a tool for the recognition of the partitive relation in sentences. For this purpose, we combine the descriptive notation, represented in FrameNet, with information from WordNet and the SIMPLE ontology, which are lexical resources that assume the Generative Lexicon Theory (PUSTEJOVSKY, 1996). The creation of an ontological database (of the entities and their parts), as a result of the effort of annotation of frames of a language, helped the recognition of the partitive relation in various sentences and promoted its readability by machines, through the explicitness of the semantic types of the elements linked by partitive relation.