Decipher NER a Decipher IE
OTRUSINA, L. SMRŽ, P. SZNAPKA, J. ŠAFÁŘ, M.
Named Entity Recognizer and Classifier takes care of identification of relevant entities - names of people, places, artworks, etc. The matching of candidate terms is realized by means of the FSA technology. The data for each particular category is stored in a separate file. A joint list is compiled into a FSA that includes all terms together with their subcategorization. The disambiguation phase takes into account a context in which a particular NE candidate string appeared decides whether it really correspond to an entity in question. Relation extractor identifies textual fragments that express a particular relation between two or more entities (or values). It can work on a simple PoS-tagged data but it can also ask the Parser package to analyse the particular piece of text in question. Relation triggers - specific expressions that signal that a relation holds - form key elements of the identification. A CRF classifier is further employed to mark beginnings and ends of particular elements appearing in the relation.
Decipher, NER, IE