Relation Extraction with Sentence Simplification Process and Entity Information

Graph-based Knowledge Bases (KBs) are composed of relational facts that can be perceived as two entities, called head and tail, linked via a relation. Processes of constructing KBs, i.e., populating them with such facts, as well as revising and updating them are of special interest. These should be performed automatically, especially in the case when the main sources of facts are textual documents. For this reason, a task of Relation Extraction (RE), i.e., predicting a relation that links two entities mentioned in a sentence, is one of the most important activities. Using RE processes, new relational facts can be extracted, and KBs can be built and updated using unstructured information. In this paper, we propose a novel procedure for RE. It is based on a sentence distilling technique that works on dependency trees and removes noisy tokens from sentences while preserving the most relevant and useful ones. In addition, the proposed procedure utilizes information about types of linked entities, it means types of relations’ heads and tails. Our neural network model using processed and new input information is evaluated on the widely used NYT dataset and compared to other state-of-the-art RE methods. Experimental results show the effectiveness of the proposed procedure against other methods.