Method of processing triple data, method of training triple data processing model, device, and medium

Patent metadata

Field	Value
Publication number	US-12339919-B2
Application number	US-202217934876-A
Country	US
Kind code	B2
Filing date	Sep 23, 2022
Priority date	Sep 23, 2021
Publication date	Jun 24, 2025
Grant date	Jun 24, 2025

Abstract

The present disclosure provides a method of processing triple data, a method of training a triple data processing model, an electronic device, and a storage medium. A specific implementation solution includes: performing a triple data extraction on text data to obtain a plurality of field data; normalizing the plurality of field data to determine target triple data, wherein the target triple data contains entity data, entity relationship data, and association entity data; and verifying a confidence level of the target triple data to obtain a verification result.

First claim

What is claimed is:

1. A method of processing triple data, comprising:
performing a triple data extraction on text data to obtain a plurality of field data;
normalizing the plurality of field data to determine target triple data, wherein the target triple data contains entity data, entity relationship data, and association entity data; and
verifying a confidence level of the target triple data to obtain a verification result,
wherein the verifying the confidence level of the target triple data to obtain the verification result comprises:
searching for web data related to the target triple data by using the target triple data as a search term;
generating target feature data according to the web data; and
inputting the target feature data into a triple data processing model, so as to obtain the verification result,
wherein the triple data processing model is obtained by training, and the training comprises:
performing a triple data extraction on text training data to obtain a plurality of training field data;
normalizing the plurality of training field data to determine target training triple data, wherein the target training triple data contains training entity data, training entity relationship data, and training association entity data; and
training the triple data processing model by using the target training triple data and a label for the target training triple data, so as to obtain a trained triple data processing model, wherein the label indicates a confidence level of the target training triple data,
wherein the training the triple data processing model by using the target training triple data and a label for the target training triple data, so as to obtain a trained triple data processing model, comprises:
searching for sample web data related to the target training triple data by using the target training triple data as a search term; and
generating target training feature data according to the sample web data, so as to train the triple data processing model by using the target training feature data and the label to obtain the trained triple data processing model.

2. The method of claim 1 , wherein the performing the triple data extraction on the text data to obtain the plurality of field data comprises:
performing the triple data extraction on the text data to obtain a plurality of initial field data;
identifying initial field data of a target type from the plurality of initial field data;
extracting candidate data from the text data responsive to the initial field data of the target type conforming to a predetermined rule; and
modifying the initial field data of the target type according to the candidate data, so as to obtain the plurality of field data.

3. The method of claim 1 , wherein the normalizing the plurality of field data to determine the target triple data comprises:
clustering field data of a target type in the plurality of field data, so as to determine a cluster; and
sorting field data in the cluster according to a number of words or a data source, and determining top field data as a target field of the target type, so as to obtain the target triple data.

4. The method of claim 1 , wherein the target feature data comprises at least one selected from: a temporal feature related to the web data, or an attribute feature related to the target triple data in the web data.

5. The method of claim 1 , further comprising:
preprocessing initial text data to obtain the text data at a sentence level,
wherein the preprocessing comprises at least one selected from: selecting, sentence segmenting, or format unifying.

6. The method of claim 1 , wherein the target training feature data comprises at least one selected from: a temporal feature related to the sample web data, or an attribute feature related to the target training triple data in the sample web data.

7. A non-transitory computer-readable storage medium having computer instructions therein, wherein the computer instructions are configured to cause a computer to implement the method of processing the triple data of claim 1 .

8. An electronic device, comprising:
at least one processor; and
a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions, when executed by the at least one processor, cause the at least one processor to implement a method of processing triple data, comprising operations of:
performing a triple data extraction on text data to obtain a plurality of field data;
normalizing the plurality of field data to determine target triple data, wherein the target triple data contains entity data, entity relationship data, and association entity data; and
verifying a confidence level of the target triple data to obtain a verification result,
wherein the instructions, when executed by the at least one processor, cause the at least one processor further to implement operations of:
searching for web data related to the target triple data by using the target triple data as a search term;
generating target feature data according to the web data; and
inputting the target feature data into a triple data processing model, so as to obtain the verification result,
wherein the triple data processing model is obtained by training, and the training comprises:
performing a triple data extraction on text training data to obtain a plurality of training field data;
normalizing the plurality of training field data to determine target training triple data, wherein the target training triple data contains training entity data, training entity relationship data, and training association entity data; and
training the triple data processing model by using the target training triple data and a label for the target training triple data, so as to obtain a trained triple data processing model, wherein the label indicates a confidence level of the target training triple data,
wherein the training the triple data processing model by using the target training triple data and a label for the target training triple data, so as to obtain a trained triple data processing model, comprises:
searching for sample web data related to the target training triple data by using the target training triple data as a search term; and
generating target training feature data according to the sample web data, so as to train the triple data processing model by using the target training feature data and the label to obtain the trained triple data processing model.

9. The electronic device of claim 8 , wherein the instructions, when executed by the at least one processor, cause the at least one processor further to implement operations of:
performing the triple data extraction on the text data to obtain a plurality of initial field data;
identifying initial field data of a target type from the plurality of initial field data;
extracting candidate data from the text data responsive to the initial field data of the target type conforming to a predetermined rule; and
modifying the initial field data of the target type according to the candidate data, so as to obtain the plurality of field data.

10. The electronic device of claim 8 , wherein the instructions, when executed by the at least one processor, cause the at least one processor further to implement operations of:
clustering field data of a target type in the plurality of field data, so as to determine a cluster; and
sorting field data in the cluster according to a number of words or a data source, and determining top field data as a target field of the target type, so as to obtain the target triple data.

11. The electronic device of claim 8 , wherein

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Wang Zhaoji
• Huang Fang
• Jiang Ye
• Shi Yabing
• Chai Chunguang
• Zhu Yong

Classifications

• G06F40/226

  Validation · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F16/355

  Creation or modification of classes or clusters · CPC title