Abstract

The aim of this paper is to present a proposal of implicitness typology. The theoretical model we propose is compliant with neo-Gricean pragmatics and is explicitly designed to cover instances of offensive language on social media. The implicitness framework we propound has been empirically verified by means of a corpus-assisted analysis and computational method of word embeddings (Word2Vec and FastText), which, in principle, have supported the schema explicated here. This taxonomy is potentially applicable to the ontology of offensiveness and, thus, to NLP-based research; in particular, it can be useful for automatic detection of implicit offensive language on social media.