Pragmatic Information Extraction in Brazilian Portuguese Documents

Abstract

The volume of published data in the Web has been increasing, and a great amount of those data is available in a natural language format. Manually analyzing each document is a time-consuming and tedious task. Thus, Open IE area emerges to help the extraction of semantic relationships in a large number of texts written in a natural language from different domains. Although a semantic analysis does not guarantee complete accuracy in extracting relations, a pragmatic analysis becomes important on Open EI to identify additional meanings (unsaid) that goes beyond semantics in a text. Our work developed a method for Open Information Extraction to extract relations from texts written in Portuguese in a first pragmatic level. We stated that a first pragmatic level deals with inferential, contextual and intentional aspects. We evaluate our approach, and our results outstand the most relevant related work on comparing accuracy and minimality measures.