Knowledge Enabled Relation Extraction

Abstract

Relation extraction is the task of extracting relationships from input text, where input can be a sentence, document, or multiple documents. This task has been popular for decades and is still of keen interest. Various techniques have been proposed to solve the relation extraction problem, among which the most popular are using distant supervision, deep learning-based models, reasoning-based models, and transformer-based models. We propose three approaches (named ReOnto, DocRE-CLip, and KDocRE) for relation extraction from text at three levels of granularity (sentence, document and across documents). These approaches embed knowledge in a deep learning based model to improve performance. ReOnto and DocRE-CLip have been evaluated and the source code is publicly available. We are currently implementing and evaluating KDocRE.