Research And Application Of Entity Relation Extraction Method Based On Convolutional Neural Network

With the rapid development of Internet technology,network resources have shown explosive growth,which has increased the difficulty of obtaining effective information.In online news,online text has the characteristics of timeliness,accuracy,and versatility,and is an important source of information acquisition.Named entity recognition and entity relationship extraction fundamentally solve the problem of classification of the relationship between the target entity and the entity in the text.It can convert unstructured data into triple structured data for storage.The core basic process of the domain knowledge map also has a strong guiding significance for people's future actions and decisions.For the two tasks of entity recognition and relationship extraction,this paper first designs a word embedding module that combines traditional word vectors and BERT character vectors,introduces various levels of vector information,and then builds an entity recognition model based on ID-CNN-CRF.In order to identify the key types of entities in the relevant sports news texts of the 2022 Beijing Winter Olympics,a sports news entity recognition data set was constructed by manual annotation,and comparative experiments were carried out with other models.Then,on the basis of the entity recognition model,the entity relationship is jointly extracted through joint learning,which solves the problem of error propagation and entity redundancy in the pipeline-based extraction process,and introduces an attention mechanism to improve the extraction effect.Finally,aiming at the problem of entity relationship overlap,the multi-classification problem of entity recognition is transformed into a multi-label problem through multi-head selection,which can select multiple relationships that exist between an entity and multiple other entities.Finally,the effectiveness of the model is verified on the open source relation extraction data set.