An Entity Relation Extraction Algorithm Incorporating Multi-Attention Mechanisms and Remote Supervision

Entity-relation extraction is one of the important tasks in information extraction, which is widely used in natural language processing, construction of knowledge graph, and information reasoning work. Remote supervised algorithms utilize heuristic alignment of knowledge base with corpus, which can generate large-scale labeled corpus resources without human involvement. For the problem that noise is introduced into remote supervised algorithms leading to a decrease in the recognition rate, this paper introduces a multi-level attention mechanism into remote supervised learning algorithms, using word attention to strengthen the role of key words related to entity relations, and using sentence attention to further emphasize the role of key sentences; for the problem that entity pairs may store more than one relation, relational attention is invoked to find a relation in sentence packages that have the same entity pairs The exact expression of the relationship is found in the package of sentences with the same entity pairs. Experiments are conducted on the Freebase+NYT corpus. The results indicate that the algorithm designed in this paper is significantly improved compared to the classical algorithms.