Authors: MEHMET AYDAR, ÖZGE BOZAL, FURKAN ÖZBAY
Abstract: Neural relation extraction discovers semantic relations between entities from unstructured text using deeplearning methods. In this study, we make a clear categorization of the existing relation extraction methods in termsof data expressiveness and data supervision, and present a comprehensive and comparative review. We describe theevaluation methodologies and the datasets used for model assessment. We explicitly state the common challenges inrelation extraction task and point out the potential of the pretrained models to solve them. Accordingly, we investigateadditional research directions and improvement ideas in this field.
Keywords: Neural relation extraction, deep learning, pretrained model, distant supervision
Full Text: PDF