An entity identification method and related equipment

Abstract

The embodiment of the invention discloses an entity identification method and related equipment, and the method comprises the steps: firstly, obtaining a plurality of annotation corpora, and enablingeach annotation corpus in the plurality of annotation corpora to carry annotation information; Establishing a hypergraph model according to a preset entity labeling rule; Determining a labeling path diagram corresponding to each labeling corpus according to the labeling information and an entity labeling rule, and establishing a to-be-trained model according to the hypergraph model and a preset neural network model; And finally, inputting the labeled path diagram into a to-be-trained model for training to obtain an entity recognition model, and recognizing at least one named entity in the input corpus according to the entity recognition model. By adopting the embodiment of the invention, the entity of the nested structure can be effectively identified, so that the accuracy of entity identification and entity extraction is improved.

Description

technical field [0001] The present invention relates to the technical field of information processing, in particular to an entity recognition method and related equipment. Background technique [0002] In the era of information explosion, how to quickly and effectively extract the required information from massive data has become a hot research topic, which has led to the research on natural language processing. For a long time, the task of entity extraction has been widely concerned in the field of natural language processing. It is the pre-step of many natural language processing tasks, so its performance also directly affects the performance of downstream natural language processing tasks, such as entity connection and entity relationship. Classification, knowledge graph reasoning, etc. Among them, entities are named entities, which refer to the names of people, institutions, places, and all other entities identified by names in natural language. More extensive entities ...

AI Technical Summary

Problems solved by technology

At present, for entity extraction tasks, the mainstream extraction models are conditional random field models and neural network-conditional random field models. Such models cannot directly handle nested structures, and can only complete nested entity recognition by superimposing multiple models. However, the way of superimposing multiple models will not be able to effectively capture the dependencies between entities because each conditional random field model is independent of each other, resulting in poor performance of entity recognition and low accuracy of entity extraction

Images