Research On Financial Relationship Extraction Method Based On Trigger Words And Dependency Syntax

With the popularization and development of the Internet,we are in an era of information explosion,text data as the main recorder of information has exploded exponentially,overwhelming complex information has affected the acquisition of knowledge,so how to extract useful structured information from unstructured information is particularly urgent and needed.As the key in information extraction,the relationship extraction task realizes the extraction of "subjectrelationship-object" triplet knowledge from unstructured text,which is used for the construction of knowledge graph and the retrieval of auxiliary information.It provides knowledge support for future business applications in professional fields.With the development of financial technology and digitalization,financial knowledge is becoming more and more valuable,and the need for knowledge utilization is becoming more and more urgent,so the realization of relationship extraction in the financial field is very important.Starting from the analysis of financial text characteristics and the construction of financial ontology,this study proposes a targeted relationship extraction method that integrates external knowledge semantic similarity features.Combining with the relationship triggered vocabulary,an unsupervised semantic matching task and similar feature extraction method are innovatively designed,which improves the coding effect in the financial field in clustering and visualization applications.The effect of financial relationship extraction is significantly improved in small sample scenarios,The F1 value has increased by more than ten percentage points.In order to solve the problem of relationship overlap in financial relationship extraction,this study adopts a syntactic structure modeling approach and designs a dependency syntax tree pruning strategy based on a combination of hardware and software,effectively removing redundant information in long sentences.For the first time,it proposes to integrate entity types and syntactic types into heterogeneous graph convolutional neural networks(HGCN),effectively integrating multi-dimensional information such as structure and semantics,The relationship extraction algorithm has increased the F1 value by about one percentage point on the financial field dataset.This study transforms algorithm research into application results,achieving the design of financial relationship extraction algorithm system,the construction process of financial knowledge graph,and the design of intelligent question answering application functions,further exploring the value of financial applications.