Feature Enhancement Based Joint Extraction of Web Novel Entity Relationships

Abstract

In an era characterized by constant advancements in computer science, web novels represent an extensive and intricate form of text that presents unique challenges for automated processing. This investigation aims to address the issues associated with the time-intensive, laborious, and error-prone nature of text processing within web novels. It presents a novel joint entity-relationship extraction model that is enhanced by various features. By leveraging a combination of computer vision and natural language processing techniques, the extraction of named entities and relationships is modeled in a unified framework to optimize text feature mining. The employment of bidirectional long-short term memory networks and multi-layer perceptron equips the model with the capability to effectively extract entity relationships from web novels comprehensively. Experimental outcomes indicate that the model achieves an F1 score of 72.4%, marking a notable enhancement over traditional pipelined models. This study offers robust tools and methodologies for computers to process extensive and complex textual data, further integrates computer vision with natural language processing, and broadens the potential applications within the domain of information processing.