US 12,333,239 B2
Copy generation method and apparatus, and electronic device, storage medium and program
Liang Xu, Beijing (CN); Xin Wang, Beijing (CN); and Yang He, Beijing (CN)
Assigned to Beijing Wodong Tianjun Information Technology Co., Ltd., Beijing (CN)
Appl. No. 18/251,055
Filed by Beijing Wodong Tianjun Information Technology Co., Ltd., Beijing (CN)
PCT Filed Oct. 29, 2021, PCT No. PCT/CN2021/127421
§ 371(c)(1), (2) Date Apr. 28, 2023,
PCT Pub. No. WO2022/095798, PCT Pub. Date May 12, 2022.
Claims priority of application No. 202011219419.8 (CN), filed on Nov. 4, 2020.
Prior Publication US 2023/0385535 A1, Nov. 30, 2023
Int. Cl. G06F 40/166 (2020.01); G06F 40/106 (2020.01); G06F 40/10 (2020.01); G06Q 30/0601 (2023.01)
CPC G06F 40/166 (2020.01) [G06F 40/106 (2020.01); G06F 40/10 (2020.01); G06Q 30/0627 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A method for generating a copy, comprising:
acquiring first attribute data of a commodity;
determining first key attribute data of the commodity based on a first copy generation model trained in advance, wherein the first key attribute data represents a part of attribute data of the first attribute data;
obtaining a first candidate copy set of the commodity according to the first key attribute data, wherein the first candidate copy set represents a set of at least one commodity copy; and
screening candidate copy data according to a quality determination rule to determine a target commodity copy, wherein the candidate copy data comprises one or more commodity copies in the first candidate copy set;
wherein screening the candidate copy data according to the quality determination rule to determine the target commodity copy comprises:
after obtaining the first attribute data of the commodity, inputting the first attribute data into at least two copy generation models to obtain a second candidate copy set of the commodity, wherein the at least two copy generation models comprises the first copy generation model; and
screening the candidate copy data according to the quality determination rule, wherein the candidate copy data comprises commodity copies in the second candidate copy set;
wherein the quality determination rule comprises:
screening quality of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set based on a duplication degree, wherein the duplication degree represents a degree of duplication between different copy descriptions in each of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set;
screening the quality of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set based on consistency, wherein the consistency represents a consistency degree between the attribute data of each of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set and the first attribute data;
screening the quality of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set based on a perplexity, wherein the perplexity represents a clarity degree of each copy description of each of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set; and
screening the quality of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set based on an attribute coverage degree, wherein the attribute coverage degree represents a degree of coverage of the first attribute data in each of the one or more commodity copies in the first candidate copy set and the commodity copies in the second candidate copy set.