Anotação Semântica Híbrida: Anotação Baseada em Regras e Manual do Open American National Corpus com Ontologias de Nível topo

Resumo

Natural language processing still faces the challenge of getting machines to understand the meaning expressed by words that occur in a sentence. Semantic annotation helps in this process by adding metadata that attaches meaning to lexemes. There are several semantic aspects that can be annotated, such as function, semantic role and ontological categories. Top-level ontological categories add information about the nature of the concept denoted by the lexeme and allow eliminating ambiguities. The proposed work is a hybrid semantic annotation approach based on top-level ontologies applied to an American English Corpus. The research is divided into two annotation steps, both using the top-level categories of Schema.org as annotation labels. In the first step a rules-based annotator is created, and
in the second step a manual annotation is made for correction and addition of labels in the corpus annotated by rule annotator. The contribution of this work is the generation of an annotated corpus that can be used in the training of automatic annotators.