Building Knowledge Graph using Pre-trained Language Model for Learning Entity-aware Relationships

Relations exhibited among entities from textual content can be a potential source of information for any business domain. This paper encompasses a wholesome approach to mine entity-relation and building knowledge graph from textual documents. The paper concentrates on two approaches to classify directional entity relations. We build on extending pretrained language model i.e. BERT for text classification along-side providing entity and directionality information as input making it entity-aware BERT classifier. We also did ablation studies of presented model in terms of various ways of providing entity information on the learning capabilities of model. We demonstrate the end to end pipeline for building an entity-relation extraction system in a business application. The techniques proposed in the paper are also evaluated against SemEval-2010 Task 8, a popular relation classification dataset. The experimental results demonstrate that learning entity-aware relations through language models outperforms almost all the previous state-of-the-art (SOTA) models.