Neural zero-sample fine-grained entity classification method based on type attention

Abstract

The invention discloses a neural zero-sample fine-grained entity classification method based on type attention, and the method specifically comprises the steps: calculating a target entity text representation vector based on a word vector corresponding to each word in a target entity text; obtaining a basic context vector of the target entity text based on the context word vectors corresponding tothe words on the two sides of the target entity text in the sentence; constructing an entity type vector based on the type corresponding to each entity in the target entity text; based on the targetentity type vector and the basic context vector, calculating related attention values corresponding to the words on the two sides respectively; calculating a related context vector of the target entity text based on the related attention value and the basic context vector; combining the entity representation vector and the related context vector of the target entity text, obtaining the representation vector of the whole sentence, obtaining the score of the target entity text belonging to each given category by using the created entity type classifier model, thus effectively improving the effect of entity fine-grained classification.

Description

technical field [0001] The invention belongs to the technical field of information processing, and in particular relates to a neural zero-sample fine-grained entity classification method based on type attention. Background technique [0002] The person in the entity classification is the semantic type that identifies the entity text, such as person name, place name, organization name, etc. The person helps to locate entities in the text precisely, which is of great significance for many other tasks in natural language processing. [0003] Entity recognition is a key basic task in the field of information extraction. The output of entity recognition can be used as the input features of many downstream natural language processing tasks to improve the performance of downstream tasks, such as relationship extraction, question answering, topic models, etc. [0004] At present, researchers have found that finer-grained entity types can greatly improve downstream tasks, so resear...

AI Technical Summary

Problems solved by technology

Due to the time-consuming and labor-intensive nature of manual features, and the strong limitations of the application range, researchers have used deep learning to replace traditional manual features in recent years.

On the other hand, in many cases, the training model requires a large amount of labeled data, and the cost of manually labeled data is very high, and entity classification and recognition with few samples or even zero samples has become a research topic.

In addition, due to the large number of fine-grained entity types and one entity mention will have multiple fine-grained entity types, the sequence annotation model used for coarse-grained entity recognition is not suitable for end-to-end fine-grained entity recognition, which is also a problem that needs to be solved

Images