Document-level Relationship Extraction Based on Evidence Enhancement

A document level relationship extraction method based on evidence augmentation is proposed to address the performance shortcomings of existing methods in processing long texts and cross sentence reasoning tasks. Extracting documents as evidence sentences to enhance the model's understanding of relationship extraction, incorporating CNN to obtain fragment level features related to entities, and obtaining richer semantic information of documents. By fusing evidence sentences with the original document, information redundancy in long document processing can be reduced, enabling the model to better understand complex relationship structures and diverse relationship types. The model can effectively learn in the absence of a large amount of annotated data, reducing the amount of text processing. The experimental results show that this method achieves a certain improvement in accuracy in document level relationship extraction tasks.