A systematic mapping study on open information extraction

Abstract

Open information extraction (Open IE) is a task for extracting relationship triples in plain texts without previously determining these relationships. The Open IE systems are generally applied to solutions on the web-scale such improving question answering systems, ontology constructions, document filtering and clustering. Since 2007, within the first Open IE system TEXTRUNNER, other related works have been proposed in this area. Despite other secondary studies on Open IE, useful information available to initiate new research in the area is limited. Thus, we propose a review of the literature in Open IE by a systematic mapping study. We have retrieved 2484 articles about Open IE in Science Direct, IEEE Xplore, ACM Digital Library, Scopus and Google Scholar databases. Among them, 2411 were filtered by exclusion criteria proposed in our systematic mapping protocol. The remaining 73 papers represent the state-of-the-art from the past seven years. Different researchers have proposed important contributions and have pointed out some open problems for Open IE. As a result, we summarized these contributions and identified significant gaps that could be envisioned as future works.