US 12,443,804 B2
Method, apparatus and storage medium for training natural language processing model
Boran Jiang, Beijing (CN); Chao Ji, Beijing (CN); Hongxiang Shen, Beijing (CN); Zhenzhong Zhang, Beijing (CN); Ge Ou, Beijing (CN); Chuqian Zhong, Beijing (CN); Shuqi Wei, Beijing (CN); and Pengfei Zhang, Beijing (CN)
Assigned to Beijing BOE Technology Development Co., Ltd., Beijing (CN); and BOE Technology Group Co., Ltd., Beijing (CN)
Appl. No. 18/031,511
Filed by Beijing BOE Technology Development Co., Ltd., Beijing (CN); and BOE Technology Group Co., Ltd., Beijing (CN)
PCT Filed Mar. 8, 2022, PCT No. PCT/CN2022/079771
§ 371(c)(1), (2) Date Apr. 12, 2023,
PCT Pub. No. WO2023/168601, PCT Pub. Date Sep. 14, 2023.
Prior Publication US 2024/0370668 A1, Nov. 7, 2024
Int. Cl. G06F 40/58 (2020.01)
CPC G06F 40/58 (2020.01) 19 Claims
OG exemplary drawing
 
1. A method, comprising:
obtaining a sample text of natural language;
determining one or more triples in the sample text, wherein each of the triples comprises two entities in the sample text and a relation between the two entities;
processing the sample text based on the triples to obtain one or more knowledge fusion vectors; and
training a natural language processing model by inputting the knowledge fusion vectors into the natural language processing model to obtain a target model;
wherein the natural language processing model comprises one or more feature extraction layers configured to:
perform linear transformation on the knowledge fusion vectors to obtain at least two knowledge fusion matrices comprising a first knowledge fusion matrix and a second knowledge fusion matrix;
determine an association matrix according to the first knowledge fusion matrix, wherein the association matrix is configured to represent association relation information between one or more entities in the sample text with the sample text; and
determine a weight matrix according to the second knowledge fusion matrix and the association matrix, wherein the weight matrix is configured to represent weight information between the entities with the sample text.