Entity linking method, entity linking device, electronic equipment and medium

Abstract

The embodiment of the invention provides an entity linking method, an entity linking device, electronic equipment and a medium, belongs to the technical field of multimedia, and aims to improve the performance of entity linking. According to the embodiments of the invention, the method comprises the steps of: receiving a to-be-identified text input by a user, determining a to-be-identified entityword in the to-be-identified text, and searching candidate entry information associated with the to-be-identified entity word from a preset knowledge base; inputting the to-be-identified text, the to-be-identified entity word and the candidate entry information into an entity linking model; encoding the to-be-identified text and the to-be-identified entity word by an encoding part of the entity linking model to obtain the feature vector of the to-be-identified text, encoding each piece of candidate entry information associated with the to-be-identified entity word to obtain the feature vectorof each piece of candidate entry information, calculating the association degree between the feature vector of the to-be-identified text and the feature vector of each piece of candidate entry information, outputting the association degree, and then determining the candidate entry information with the highest association degree as entry information linked to the to-be-identified entity word.

Description

technical field [0001] The present invention relates to the field of multimedia technology, in particular to a method, device, electronic equipment and medium for entity linking. Background technique [0002] Entity linking refers to linking an entity word in a text to determine an entity that can accurately express the semantics of the entity word linked to the entity word according to the context semantics of the entity word. For example, the word "apple" has multiple meanings, for example, it can refer to the fruit apple, the Apple company, and it can also represent a movie named Apple. For a text "Apple in the supermarket is very sweet", the entity link in the text can be determined The actual meaning of "apple" is the fruit apple. [0003] The entity linking method in the related art is to input the text into the Logistic Regression (LR) model, analyze the text by the LR model, and determine the target entity word corresponding to the language feature matching the enti...

AI Technical Summary

Problems solved by technology

However, it is difficult to collect enough language features of text data manually. When the entity link of a certain entity word is carried out through the logistic regression model, if the language features of the entity word are not collected when the logistic regression model is established, or the collected Inaccurate language features will make it impossible to determine the target entity word corresponding to the entity word, or cause the determined target entity word to be inaccurate

It can be seen that the performance of entity linking through the logistic regression model in the prior art is poor

Images