US 12,431,137 B2
Method of detecting a transcription error in speech recognition corpus and device for the same
Myoung Wan Koo, Seoul (KR); and Jeong Pil Lee, Seoul (KR)
Assigned to SOGANG UNIVERSITY RESEARCH & BUSINESS DEVELOPMENT FOUNDATION, Seoul (KR)
Filed by SOGANG UNIVERSITY RESEARCH & BUSINESS DEVELOPMENT FOUNDATION, Seoul (KR)
Filed on Dec. 7, 2023, as Appl. No. 18/532,770.
Claims priority of application No. 10-2022-0173084 (KR), filed on Dec. 12, 2022.
Prior Publication US 2024/0194203 A1, Jun. 13, 2024
Int. Cl. G10L 15/26 (2006.01); G10L 15/01 (2013.01); G10L 15/22 (2006.01)
CPC G10L 15/26 (2013.01) [G10L 15/01 (2013.01); G10L 15/22 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A method of detecting a transcription error in speech recognition corpus includes following steps:
(a) receiving the speech recognition corpus including a speech file and a text label for the speech file;
(b) performing speech recognition on the speech file in the speech recognition corpus using a speech recognition model and converting the speech recognition result into a text;
(c) extracting a performance evaluation index for the speech recognition result of the speech recognition model;
(d) extracting a PPL(s2) for the text label of the speech recognition corpus and a PPL(s1) for the text according to the speech recognition result using a language model; and
(e) detecting the transcription error in the text label of the speech recognition corpus using the extracted performance evaluation index and the extracted PPL(s2) and PPL(s1).