Information processing method, related device, and computer storage medium

Patent metadata

Field	Value
Publication number	US-11734341-B2
Application number	US-202017070738-A
Country	US
Kind code	B2
Filing date	Oct 14, 2020
Priority date	Jun 15, 2018
Publication date	Aug 22, 2023
Grant date	Aug 22, 2023

Abstract

An information processing method is disclosed. The method includes: obtaining initial feature information of to-be-queried information; performing feature processing on the initial feature information of the to-be-queried information, to obtain target feature information of the to-be-queried information, where the feature processing includes at least feature dimension processing used to process a data dimension of the initial feature information of the to-be-queried information to be a preset dimension; performing a first retrieval operation on m pieces of candidate information based on the target feature information of the to-be-queried information, to obtain a first retrieval result; and performing a second retrieval operation on the first retrieval result based on the initial feature information of the to-be-queried information, to obtain result information corresponding to the to-be-queried information. The method can quickly and efficiently implement information retrieval.

First claim

What is claimed is:

1. An information processing method for use by a terminal device including a processor comprising:
obtaining by the processor initial feature information of to-be-queried information;
performing by the processor feature processing on the initial feature information of the to-be-queried information to obtain target feature information of the to-be-queried information, wherein the feature processing comprises at least feature dimension processing, and the feature dimension processing is used to process a data dimension of the initial feature information of the to-be-queried information to be a preset dimension, wherein the feature processing further comprises feature quantization processing, the feature quantization processing is used to perform feature quantization on second feature information, wherein the second feature information is feature information obtained after the feature dimension processing is performed on the initial feature information of the to-be-queried information and the target feature information of the to-be-queried information is quantized feature information;
performing by the processor a first retrieval operation on m pieces of candidate information based on the target feature information of the to-be-queried information to obtain a first retrieval result, wherein m is a positive integer; and
performing by the processor a second retrieval operation on the first retrieval result based on the initial feature information of the to-be-queried information to obtain result information corresponding to the to-be-queried information.

2. The method according to claim 1 , wherein each piece of candidate information comprises initial feature information of the candidate information and target feature information of the candidate information, and a data dimension of the target feature information of the candidate information is the same as a data dimension of the target feature information of the to-be-queried information;
wherein, the first retrieval operation comprises:
retrieving target feature information of the m pieces of candidate information based on the target feature information of the to-be-queried information to obtain the first retrieval result, wherein the first retrieval result comprises n pieces of candidate information, and n is a positive integer less than m; and
the second retrieval operation comprises:
retrieving initial feature information of the n pieces of candidate information based on the initial feature information of the to-be-queried information to obtain the result information corresponding to the to-be-queried information.

3. The method according to claim 1 , wherein the feature processing further comprises feature equalization processing, the feature equalization processing is used to perform feature equalization on first feature information, equalized first feature information is distributed according to a preset rule, and the first feature information is feature information obtained after the feature dimension processing is performed on the initial feature information of the to-be-queried information.

4. The method according to claim 3 , wherein the feature processing further comprises feature quantization processing, the feature quantization processing is used to perform feature quantization on third feature information, wherein the third feature information is feature information obtained after the feature equalization processing is performed on the initial feature information of the to-be-queried information.

5. The method according to claim 1 , wherein the target feature information of each piece of candidate information is non-quantized feature information, each piece of candidate information further comprises index information of the candidate information, and the index information is quantized feature information; and
before the performing a first retrieval operation on m pieces of candidate information based on the target feature information of the to-be-queried information, the method further comprises:
obtaining the target feature information of the m pieces of candidate information based on index information of the m pieces of candidate information.

6. A terminal device comprising:
at least one processor, and a memory storing instructions, which when executed by the at least one processor, cause the at least one processor to perform operations comprising:
obtaining initial feature information of to-be-queried information;
performing feature processing on the initial feature information of the to-be-queried information to obtain target feature information of the to-be-queried information, wherein the feature processing comprises at least feature dimension processing, and the feature dimension processing is used to process a data dimension of the initial feature information of the to-be-queried information to be a preset dimension, wherein the feature processing further comprises feature quantization processing, the feature quantization processing is used to perform feature quantization on second feature information, wherein the second feature information is feature information obtained after the feature dimension processing is performed on the initial feature information of the to-be-queried information and the target feature information of the to-be-queried information is quantized feature information;
performing a first retrieval operation on m pieces of candidate information based on the target feature information of the to-be-queried information to obtain a first retrieval result, wherein m is a positive integer; and
performing a second retrieval operation on the first retrieval result based on the initial feature information of the to-be-queried information to obtain result information corresponding to the to-be-queried information.

7. The terminal device according to claim 6 , wherein each piece of candidate information comprises initial feature information of the candidate information and target feature information of the candidate information, and a data dimension of the target feature information of the candidate information is the same as a data dimension of the target feature information of the to-be-queried information;
wherein, the first retrieval operation comprises:
retrieving target feature information of the m pieces of candidate information based on the target feature information of the to-be-queried information to obtain the first retrieval result, wherein the first retrieval result comprises n pieces of candidate information, and n is a positive integer less than m; and
the second retrieval operation comprises:
retrieving initial feature information of the n pieces of candidate information based on the initial feature information of the to-be-queried information to obtain the result information corresponding to the to-be-queried information.

8. The terminal device according to claim 6 , wherein the feature processing further comprises feature equalization processing, the feature equalization processing is used to perform feature equalization on first feature information, equalized first feature information is distributed according to a preset rule, and the first feature information is feature information obtained after the feature dimension processing is performed on the initial feature information of the to-be-queried information.

9. The terminal device according to claim 8 , wherein the feature processing further comprises feature quantization processing, the feature quantization processing is used to perform feature quantization on third feature information, wherein the third feature information is feature information obtained after the feature equalization processing is performed on the initial feature information of the to-be-quer

Assignees

• Huawei Tech Co Ltd

Inventors

• Wu Huajingling
• Han Tingrui
• Ni Ningxi
• Wang Lele
• Sheng Zhenli

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06F18/213

  Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods · CPC title

• G06F16/583 Primary

  using metadata automatically derived from the content · CPC title

• G06F16/906

  Clustering; Classification · CPC title

• G06F16/908

  using metadata automatically derived from the content · CPC title