Title:
MULTILINGUAL SPEECH RECOGNITION METHOD AND APPARATUS, AND ELECTRONIC DEVICE
Document Type and Number:
WIPO Patent Application WO/2021/179701
Kind Code:
A1
Abstract:
A multilingual speech recognition method and apparatus, and an electronic device, relating to the technical field of speech recognition. The multilingual speech recognition method comprises: acquiring a target speech to be recognized (S1); calling a pre-trained acoustic model and a pre-trained multilingual language model to decode the target speech, and acquiring a recognition result search grid of the target speech (S2); calling a plurality of pre-trained monolingual language models to respectively re-score the recognition result search grid, obtaining, by screening, the candidate recognition result of each corresponding language, and determining the probability that the candidate recognition result is a target recognition result of the target speech (S3); and ranking the candidate recognition results in descending order of probability, and obtaining, by screening, the target recognition result from a preset number of candidate recognition results ranked at the top (S4). According to the present solution, the difficulty of multilingual speech recognition can be reduced. Similarly, the solution can be applied to online consultation in digital healthcare.
Inventors:
LIU BOQING (CN)
WANG JIANZONG (CN)
ZHANG ZHIYONG (CN)
CHENG NING (CN)
WANG JIANZONG (CN)
ZHANG ZHIYONG (CN)
CHENG NING (CN)
Application Number:
PCT/CN2020/134543
Publication Date:
September 16, 2021
Filing Date:
December 08, 2020
Export Citation:
Assignee:
PING AN TECH SHENZHEN CO LTD (CN)
International Classes:
G10L15/00
Foreign References:
| CN110895932A | 2020-03-20 | |||
| CN111369978A | 2020-07-03 | |||
| CN109817213A | 2019-05-28 | |||
| CN110491382A | 2019-11-22 | |||
| JP2018087935A | 2018-06-07 |
Attorney, Agent or Firm:
SCIHEAD IP LAW FIRM (CN)
Download PDF: