US 12,406,674 B2
Text processing method and apparatus, electronic device, and medium
Jingsheng Yang, Beijing (CN); Chunsai Du, Beijing (CN); Wenming Xu, Beijing (CN); Li Zhao, Beijing (CN); and Xiao Han, Beijing (CN)
Assigned to BEIJING BYTEDANCE NETWORK TECHNOLOGY CO., LTD., Beijing (CN)
Appl. No. 18/043,514
Filed by BEIJING BYTEDANCE NETWORK TECHNOLOGY CO., LTD., Beijing (CN)
PCT Filed Aug. 24, 2021, PCT No. PCT/CN2021/114184
§ 371(c)(1), (2) Date Feb. 28, 2023,
PCT Pub. No. WO2022/042512, PCT Pub. Date Mar. 3, 2022.
Claims priority of application No. 202010899805.X (CN), filed on Aug. 31, 2020.
Prior Publication US 2023/0326466 A1, Oct. 12, 2023
Int. Cl. G10L 15/22 (2006.01); G10L 17/02 (2013.01); G10L 17/14 (2013.01); G10L 17/20 (2013.01)
CPC G10L 17/20 (2013.01) [G10L 17/02 (2013.01); G10L 17/14 (2013.01); G10L 15/22 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A text processing method, applied to a real-time interactive scene or a non-real-time interactive scene, comprising:
acquiring target text information generated based on audio information;
determining a to-be-error-corrected word in the target text information and a target candidate replacement word corresponding to the to-be-error-corrected word; and
determining, according to the target candidate replacement word, a target replacement word corresponding to the to-be-error-corrected word, and updating the target text information based on the target replacement word;
wherein determining, according to the target candidate replacement word, the target replacement word corresponding to the to-be-error-corrected word, and updating the target text information based on the target replacement word comprises:
for each target candidate replacement word, determining a sentence to which the to-be-error-corrected word belongs and an association feature between the each target candidate replacement word and the sentence, wherein the association feature comprises a similarity between a target candidate word and the sentence;
inputting the association feature into a pre-trained feature processing model to acquire a matching degree value between the target candidate replacement word and the sentence; and
based on a matching degree value of the each target candidate replacement word, determining, from the target candidate replacement word, the target replacement word corresponding to the to-be-error-corrected word.