ATP-OIE: An Autonomous Open Information Extraction Method

This paper describes an innovative Open Information Extraction method known as ATP-OIE1. It utilizes extraction patterns to find semantic relations. These patterns are generated automatically from examples, so it has greater autonomy than methods based on fixed rules. ATP-OIE can also summon other methods, ReVerb and ClausIE, if it is unable to find valid semantic relations in a sentence, thus improving its recall. In these cases, it is capable of generating new extraction patterns online, which improves its autonomy. It also implements different mechanisms to prevent common errors in the extraction of semantic relations. Lastly, ATP-OIE was compared with other state-of-the-art methods in a well known texts database: Reuters-21578, obtaining a higher precision than with other methods.