Granted Patent B2

US 12,632,656 · App. 18/272,516 · Granted May 19, 2026

Inventors: Haoran Li (Beijing, CN); Peng Yuan (Beijing, CN)

Assignees: BEIJING WODONG TIANJUN INFORMATION TECHNOLOGY CO., LTD.; BEIJING JINGDONG CENTURY TRADING CO., LTD.

G06F40/284

Abstract

Claims (78)

1 . A text generation method, comprising:

determining a decoded-word information group set on the basis of a text to be processed, wherein the text to be processed is used for describing a specified object;

de-duplicating the decoded-word information group set to generate a candidate-word information group set;

selecting, from each candidate-word information group in the candidate-word information group set, candidate-word information that meets a target condition as target-word information, to obtain a target-word information set;

for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a convergence condition, splicing, on the basis of a preset-word list, the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a target text,

wherein the decoded-word information in the decoded-word information group set includes: decoded words and decoded-word probability values corresponding to the decoded words; and

the de-duplicating the decoded-word information group set to generate a candidate-word information group set includes:

for each decoded-word information group in the decoded-word information group set, dividing, based on the decoded words, the decoded-word information group into a duplicated decoded-word information group and a single decoded-word information group;

selecting, from the duplicated decoded-word information group, duplicated decoded-word information that meets a preset condition as target duplicated decoded-word information;

splicing the target duplicated decoded-word information with the single decoded-word information group to generate the candidate-word information group.

2 . The method of claim 1 , wherein, the determining a decoded-word information group set on the basis of a text to be processed includes:

inputting the text to be processed into a text encoder to generate an encoding hidden layer vector;

inputting the encoding hidden layer vector into a decoder to generate a decoded-word information group set.

3 . The method of claim 2 , wherein, the candidate-word information in the candidate-word information group set includes: candidate words and candidate-word probability values corresponding to the candidate words; and

the selecting, from each candidate-word information group in the candidate-word information group set, candidate-word information that meets a target condition as target-word information includes:

for each candidate-word information group in the candidate-word information group set, selecting at least one piece of candidate-word information from the candidate-word information group in descending order of the candidate-word probability values;

placing candidate-word information with a largest probability value in the at least one piece of candidate-word information, in the historical target-word information set, determining at least one piece of initial target-word information of the candidate-word information with the largest probability value, and generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information; the initial target-word information in the at least one piece of initial target-word information includes: initial target words and initial target-word probability values corresponding to the initial target words.

4 . The method of claim 2 , wherein, the for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a convergence condition, splicing, on the basis of a preset-word list, the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a target text includes:

for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a preset convergence condition, splicing the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a spliced text;

selecting, from the preset-word list, the words matching the spliced text as conjunctions;

combining the conjunctions with the spliced text to generate the target text.

5 . The method of claim 1 , wherein, the for each decoded-word information group in the decoded-word information group set, dividing, based on the decoded words, the decoded-word information group into a duplicated decoded-word information group and a single decoded-word information group includes:

for each decoded-word information group in the decoded-word information group set, in response to the decoded-word information group having other decoded-word information containing decoded words of the said decoded-word information, placing the decoded-word information and the other decoded-word information in the duplicated decoded-word information group; otherwise, placing the decoded-word information in the single decoded-word information group.

6 . The method of claim 5 , wherein, the candidate-word information in the candidate-word information group set includes: candidate words and candidate-word probability values corresponding to the candidate words; and

the selecting, from each candidate-word information group in the candidate-word information group set, candidate-word information that meets a target condition as target-word information includes:

for each candidate-word information group in the candidate-word information group set, selecting at least one piece of candidate-word information from the candidate-word information group in descending order of the candidate-word probability values;

placing candidate-word information with a largest probability value in the at least one piece of candidate-word information, in the historical target-word information set, determining at least one piece of initial target-word information of the candidate-word information with the largest probability value, and generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information; the initial target-word information in the at least one piece of initial target-word information includes: initial target words and initial target-word probability values corresponding to the initial target words.

7 . The method of claim 1 , wherein, the candidate-word information in the candidate-word information group set includes: candidate words and candidate-word probability values corresponding to the candidate words; and

the selecting, from each candidate-word information group in the candidate-word information group set, candidate-word information that meets a target condition as target-word information includes:

for each candidate-word information group in the candidate-word information group set, selecting at least one piece of candidate-word information from the candidate-word information group in descending order of the candidate-word probability values;

placing candidate-word information with a largest probability value in the at least one piece of candidate-word information, in the historical target-word information set, determining at least one piece of initial target-word information of the candidate-word information with the largest probability value, and generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information; the initial target-word information in the at least one piece of initial target-word information includes: initial target words and initial target-word probability values corresponding to the initial target words.

8 . The method of claim 7 , wherein, the generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information includes:

in response to the initial target word corresponding to the initial target-word information with the largest initial target-word probability value not belonging to the historical target-word information set, setting the initial target-word information with the largest initial target-word probability value as the target-word information, and placing the target-word information in the historical target-word information set.

9 . The method of claim 7 , wherein, the generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information includes:

in response to the initial target word corresponding to the initial target-word information with the largest initial target-word probability value belonging to the historical target-word information set, determining a duplicated-word probability difference between the initial target-word information with the largest initial target-word probability value and the historical target-word information corresponding to the historical target-word information set;

in response to the duplicated-word probability difference being greater than the initial target-word probability value of other initial target-word information corresponding to the candidate-word information, setting the initial target-word information with the largest initial target-word probability value as the target-word information, and placing the target-word information in the historical target-word information set.

10 . The method of claim 9 , wherein, the generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information includes:

in response to the initial target word corresponding to the initial target-word information with the largest initial target-word probability value not belonging to the historical target-word information set, setting the initial target-word information with the largest initial target-word probability value as the target-word information, and placing the target-word information in the historical target-word information set.

11 . The method of claim 7 , wherein, the generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information includes:

in response to the initial target word corresponding to the initial target-word information with the largest initial target-word probability value belonging to the historical target-word information set, determining a duplicated-word probability difference between the initial target-word information with the largest initial target-word probability value and the historical target-word information corresponding to the historical target-word information set;

in response to the duplicated-word probability difference being less than or equal to the initial target-word probability value of other initial target-word information corresponding to the candidate-word information, setting the initial target-word information with the largest initial target-word probability value in other initial target-word information as the target-word information, and placing the target-word information in the historical target-word information set.

12 . The method of claim 11 , wherein, the generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information includes:

in response to the initial target word corresponding to the initial target-word information with the largest initial target-word probability value not belonging to the historical target-word information set, setting the initial target-word information with the largest initial target-word probability value as the target-word information, and placing the target-word information in the historical target-word information set.

13 . The method of claim 1 , wherein, the for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a convergence condition, splicing, on the basis of a preset-word list, the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a target text includes:

for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a preset convergence condition, splicing the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a spliced text;

selecting, from the preset-word list, the words matching the spliced text as conjunctions;

combining the conjunctions with the spliced text to generate the target text.

14 . The method of claim 1 , wherein, the candidate-word information in the candidate-word information group set includes: candidate words and candidate-word probability values corresponding to the candidate words; and

the selecting, from each candidate-word information group in the candidate-word information group set, candidate-word information that meets a target condition as target-word information includes:

for each candidate-word information group in the candidate-word information group set, selecting at least one piece of candidate-word information from the candidate-word information group in descending order of the candidate-word probability values;

placing candidate-word information with a largest probability value in the at least one piece of candidate-word information, in the historical target-word information set, determining at least one piece of initial target-word information of the candidate-word information with the largest probability value, and generating target-word information corresponding to the candidate-word information based on the at least one piece of initial target-word information; the initial target-word information in the at least one piece of initial target-word information includes: initial target words and initial target-word probability values corresponding to the initial target words.

15 . The method of claim 1 , wherein, the for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a convergence condition, splicing, on the basis of a preset-word list, the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a target text includes:

for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a preset convergence condition, splicing the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a spliced text;

selecting, from the preset-word list, the words matching the spliced text as conjunctions;

combining the conjunctions with the spliced text to generate the target text.

16 . An electronic device, comprising:

at least one processor;

a storage apparatus on which at least one program are stored, and

when at least one program are executed by at least one processor, the at least one processor implement a text generation method, comprising:

determining a decoded-word information group set on the basis of a text to be processed, wherein the text to be processed is used for describing a specified object;

de-duplicating the decoded-word information group set to generate a candidate-word information group set;

selecting, from each candidate-word information group in the candidate-word information group set, candidate-word information that meets a target condition as target-word information, to obtain a target-word information set;

for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a convergence condition, splicing, on the basis of a preset-word list, the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a target text,

wherein the decoded-word information in the decoded-word information group set includes: decoded words and decoded-word probability values corresponding to the decoded words; and

the de-duplicating the decoded-word information group set to generate a candidate-word information group set includes:

for each decoded-word information group in the decoded-word information group set, dividing, based on the decoded words, the decoded-word information group into a duplicated decoded-word information group and a single decoded-word information group;

selecting, from the duplicated decoded-word information group, duplicated decoded-word information that meets a preset condition as target duplicated decoded-word information;

splicing the target duplicated decoded-word information with the single decoded-word information group to generate the candidate-word information group.

17 . A non-transitory computer-readable medium, on which a computer program is stored, wherein a text generation method is implemented when the program is executed by a processor, the text generation method, comprising:

determining a decoded-word information group set on the basis of a text to be processed, wherein the text to be processed is used for describing a specified object;

de-duplicating the decoded-word information group set to generate a candidate-word information group set;

selecting, from each candidate-word information group in the candidate-word information group set, candidate-word information that meets a target condition as target-word information, to obtain a target-word information set;

for each piece of target-word information in the target-word information set, in response to determining that the target-word information meets a convergence condition, splicing, on the basis of a preset-word list, the target-word information with a historical target-word information sequence corresponding to the target-word information, to generate a target text,

wherein the decoded-word information in the decoded-word information group set includes: decoded words and decoded-word probability values corresponding to the decoded words; and

the de-duplicating the decoded-word information group set to generate a candidate-word information group set includes:

for each decoded-word information group in the decoded-word information group set, dividing, based on the decoded words, the decoded-word information group into a duplicated decoded-word information group and a single decoded-word information group;

selecting, from the duplicated decoded-word information group, duplicated decoded-word information that meets a preset condition as target duplicated decoded-word information;

splicing the target duplicated decoded-word information with the single decoded-word information group to generate the candidate-word information group.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (19)