Name Entity Recognition And Relation Extraction For Fishery Standard Knowledge Graph Construction

Fishery modernization and fishery standardization are the development directions of digital fishery,which need to be supported by accurate standard information services,and accurate standard information services need to accurately represent the content of fishery standards.The knowledge map of fishery standards is an effective tool for the representation of fishery standard content,entity identification and relationship.Extraction is the key technology of knowledge graph construction,and the effect of entity recognition and relation extraction will directly affect the construction quality of fishery standard knowledge graph.Because the fishery standard text contains a large number of proper nouns such as fishery standard numbers and index names,and there are problems such as sparse entity samples and overlapping relationships between entities,the general entity recognition and relationship extraction methods cannot achieve effective extraction of fishery standard entities and relationships.According to the characteristics of fishery standard texts,research effective fishery standard named entity recognition and relation extraction technology.Therefore,this paper conducts research on entity recognition and entity relationship extraction for fishery standard knowledge graph construction.Specific studies are as follows.In order to solve the problem of poor effect caused by sparse corpus distribution of some entities in fishery standard named entity recognition task,a method of fishery standard named entity recognition based on multiple combination data enlargement is proposed,which combines the joint replacement algorithm based on domain dictionary,random deletion algorithm based on slot protection and random insertion algorithm based on slot protection to augment the data of corpus.First to build the "name" of aquatic products similar synonyms dictionary word dictionary and domain,through two dictionaries in aquatic product name entity and random words similar to replace and synonyms replacement,generate new sentences,to increase the number of target entities and the diversity of the sentence,and then in the case of trough point based protection of the original sentence for random delete and random insertion operation respectively,while keeping the entity and its context feature rich the diversity of the corpora,under the condition of improving the generalization ability of the model.In order to verify the effectiveness of the proposed method,several groups of comparative experiments were designed.The experimental results show that the identification accuracy,recall rate and F value of fishery standard named entity based on the method of multiple combination data amplification reach 91.73%,88.64% and 90.16%respectively,which has a good effect.The research shows that the fishery standard named entity recognition method based on multivariate combination data augmentation proposed in this paper effectively solves the problem of sparse part of the entity samples and improves the overall effect of fishery standard named entity recognition.(1)For a bad effect caused by overlapping relationship problems fishery standard entity relation extraction task,the proposed entity relation extraction based on the dual attention mechanism method.First,a sentence classification and labeling strategy is proposed to solve the problem of difficulty in labeling overlapping relations in fishery standard texts;second,a combination of dual attention mechanism and BERT-BiLSTM-CRF(Bidirectional Encoder Representations from Transformers-Bi-directional Long Short-Term Memory-Conditional Random Field)word level and sentence level attention The force mechanism layer is used to increase the weight of the target words and sentences in the paragraph,eliminate noise,and improve the accuracy of relationship extraction;Finally,a comparative experiment was designed to verify the effectiveness of the proposed method.The results show that the entity relationship extraction method based on the dual attention mechanism achieves the accuracy,recall,and F1 value on the DLOU-FSI(Fishery Standard Interaction)data set.92.67%,92.31%,92.49%.Research shows that this method can effectively solve the problem of overlapping relations in the extraction of fishery standard relations,improve the overall effect of the extraction of fishery standard entity relations,and lay a foundation for the construction of fishery standard knowledge graphs.