Abstract
Current methods to extract relational triples directly make a predictionbased on a possible entity pair in a raw sentence without depending on entityrecognition. The task suffers from a serious semantic overlapping problem, inwhich several relation triples may share one or two entities in a sentence. Inthis paper, based on a two-dimensional sentence representation, abi-consolidating model is proposed to address this problem by simultaneouslyreinforcing the local and global semantic features relevant to a relationtriple. This model consists of a local consolidation component and a globalconsolidation component. The first component uses a pixel differenceconvolution to enhance semantic information of a possible triple representationfrom adjacent regions and mitigate noise in neighbouring neighbours. The secondcomponent strengthens the triple representation based a channel attention and aspatial attention, which has the advantage to learn remote semanticdependencies in a sentence. They are helpful to improve the performance of bothentity identification and relation type classification in relation tripleextraction. After evaluated on several publish datasets, the bi-consolidatingmodel achieves competitive performance. Analytical experiments demonstrate theeffectiveness of our model for relational triple extraction and give motivationfor other natural language processing tasks.