Method, apparatus, device, and storage medium for linking entity

Patent metadata

Field	Value
Publication number	US-11727216-B2
Application number	US-202017117553-A
Country	US
Kind code	B2
Filing date	Dec 10, 2020
Priority date	Jun 9, 2020
Publication date	Aug 15, 2023
Grant date	Aug 15, 2023

Abstract

A method, apparatus, device, and storage medium for linking an entity, relates to the technical fields of knowledge graph and deep learning are provided. The method may include: acquiring a target text; determining at least one entity mention included in the target text and a candidate entity corresponding to each entity mention; determining an embedding vector of each candidate entity based on the each candidate entity and a preset entity embedding vector determination model; determining context semantic information of the target text based on the target text and each embedding vector; determining type information of the at least one entity mention; and determining an entity linking result of the at least one entity mention, based on the each embedding vector, the context semantic information, and each type information.

First claim

What is claimed is:

1. A method for linking an entity, the method comprising:
acquiring a target text;
determining at least one entity mention comprised in the target text and a candidate entity corresponding to each entity mention;
determining an embedding vector of each candidate entity, based on the each candidate entity and a preset entity embedding vector determination model;
determining context semantic information of the target text, based on the target text and each embedding vector;
determining type information of the at least one entity mention based on a context vocabulary of each entity mention; and
determining an entity linking result of the at least one entity mention, based on the each embedding vector, the context semantic information, and each type information.

2. The method according to claim 1 , wherein the entity embedding vector determination model comprises a first vector determination model and a second vector determination model, the first vector determination model representing a corresponding relationship between a description text of an entity and an embedding vector, and the second vector determination model representing a corresponding relationship between relationship information between entities and an embedding vector.

3. The method according to claim 2 , wherein the determining the embedding vector of the each candidate entity based on the each candidate entity and the preset entity embedding vector determination model, comprises:
acquiring a description text of the each candidate entity;
determining a first embedding vector of the each candidate entity based on each description text and the first vector determination model;
determining relationship information between candidate entities;
determining a second embedding vector of the each entity mention, based on the relationship information between candidate entities and the second vector determination model; and
determining the embedding vector of the each candidate entity, based on the first embedding vector and the second embedding vector.

4. The method according to claim 1 , wherein the determining context semantic information of the target text based on the target text and each embedding vector, comprises:
determining a word vector sequence of the target text; and
determining the context semantic information, based on the word vector sequence and the each embedding vector.

5. The method according to claim 4 , wherein the determining the word vector sequence of the target text, comprises:
determining an embedding vector of a candidate entity corresponding to the entity linking result, in response to acquiring the entity linking result of the at least one entity mention; and
updating the word vector sequence using the determined embedding vector.

6. The method according to claim 1 , wherein the determining type information of the at least one entity mention, comprises:
for each entity mention, occluding the entity mention in the target text; and
determining the type information of the entity mention, based on the occluded target text and a pre-trained language model.

7. The method according to claim 1 , wherein the determining the entity linking result of the at least one entity mention, based on each embedding vector, the context semantic information, and each type information, comprises:
determining the candidate entity corresponding to the each entity mention, based on the each embedding vector, the context semantic information, the each type information, and a preset learning to rank model, and using the determined candidate entity as the entity linking result of the at least one entity mention.

8. The method according to claim 1 , wherein the determining the entity linking result of the at least one entity mention, based on each embedding vector, the context semantic information, and each type information, comprises:
for each entity mention, determining a similarity between the entity mention and the each candidate entity, based on the context semantic information, an embedding vector of the entity mention, the type information of the entity mention, and a vector of the each candidate entity corresponding to the entity mention; and
determining a candidate entity having a highest similarity as the entity linking result of the entity mention.

9. The method according to claim 1 , wherein the determining the entity linking result of the at least one entity mention, based on each embedding vector, the context semantic information, and each type information, comprises:
for each entity mention, determining the entity linking result of the entity mention, based on the context semantic information and the embedding vector of the entity mention; and
verifying the entity linking result using the type information of the entity mention.

10. An electronic device for linking an entity, comprising:
at least one processor; and
a memory, communicatively connected with the at least one processor;
the memory storing instructions executable by the at least one processor, the instructions, when executed by the at least one processor, causing the at least one processor to perform operations, the operations comprising:
acquiring a target text;
determining at least one entity mention comprised in the target text and a candidate entity corresponding to each entity mention;
determining an embedding vector of each candidate entity, based on the each candidate entity and a preset entity embedding vector determination model;
determining context semantic information of the target text, based on the target text and each embedding vector;
determining type information of the at least one entity mention based on a context vocabulary of each entity mention; and
determining an entity linking result of the at least one entity mention, based on the each embedding vector, the context semantic information, and each type information.

11. The electronic device according to claim 10 , wherein the entity embedding vector determination model comprises a first vector determination model and a second vector determination model, the first vector determination model representing a corresponding relationship between a description text of an entity and an embedding vector, and the second vector determination model representing a corresponding relationship between relationship information between entities and an embedding vector.

12. The electronic device according to claim 11 , wherein the determining the embedding vector of the each candidate entity based on the each candidate entity and the preset entity embedding vector determination model, comprises:
acquiring a description text of the each candidate entity;
determining a first embedding vector of the each candidate entity based on each description text and the first vector determination model;
determining relationship information between candidate entities;
determining a second embedding vector of the each entity mention, based on the relationship information between candidate entities and the second vector determination model; and
determining the embedding vector of the each candidate entity, based on the first embedding vector and the second embedding vector.

13. The electronic device according to claim 10 , wherein the determining context semantic information of the target text based on the target text and each embedding vector, comprises:
determining a word vector sequence of the target text; and
determining the context semantic information, based on the word vector sequence and the each embedding vector.

14. The electronic device according to claim 13 , wherein the determining the wo

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Liu Zhijie
• Wang Qi
• Feng Zhifan
• Chai Chunguang
• Zhu Yong

Classifications

• G06N3/0985

  Hyperparameter optimisation; Meta-learning; Learning-to-learn · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06F17/16

  Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)} · CPC title

• G06F40/295 Primary

  Named entity recognition · CPC title