A Method of Entity Fine Classification

Abstract

The invention provides a fine classification method for an entity. The method comprises the steps of calculating an entity representation vector based on a word vector corresponding to each word in atarget entity text; acquiring a basic context vector of the target entity text based on context word vectors corresponding to words at two sides of the target entity text in a sentence; calculating knowledge base related attention values corresponding to the words at two sides based on a knowledge base representation vector and the basic context vector of the target entity text; calculating a knowledge base related context value of the target entity text based on the knowledge base related attention value and the basic context vector; combining the entity representation vector and the knowledge base related context vector of the target entity text to acquire a sentence representation vector; and acquiring a probability to which the target entity text belongs to each given category based onthe sentence representation vector by use of a constructed target classifier model. With the fine classification method for the entity, the stability of a classification model can be effectively enhanced and the effect of fine classification for the entity is effectively improved.

Description

technical field [0001] The present invention relates to the technical field of information processing, and more specifically, to a method for finely classifying entities. Background technique [0002] The task of entity classification is to identify the semantic type of entity text, such as person name, place name, organization name, etc. This task helps to accurately locate entities in text, and is of great significance to many other tasks in natural language processing. [0003] Entity fine classification is a new research direction of entity classification, which can divide entities into finer categories, such as people-athletes or place names-countries, etc. Existing fine classification methods include traditional methods involving features and deep learning methods using word vectors as features. Among them, the method of deep learning is adopted: [0004] On the one hand, consider the entity itself and the context as two independent factors, and judge the classifica...

AI Technical Summary

Problems solved by technology

However, due to the separation of entity and context and the separation of text and knowledge base in this method, the effect of fine classification of entities will be poor.

Images