Method and system for updating speech recognition resources

Abstract

The invention discloses a speech recognition resource updating method and system, wherein the method includes: receiving new speech resources within a predetermined time; performing per-update check on the new speech recognition resources in a mirror environment; if the pre-update check results detected shows that the new speech recognition sources pass the pre-update check, updating a resource file based on the new speech recognition resources in the mirror environment; using a preset full-volume test set to run a full-volume test on the updated resource file; when the recourse file passes the full-volume test, updating the new speech recognition resources to an online environment within a set period of time. The speech recognition resource updating method and system allow speech recognition resources to be updated quickly, speech service updating cost can be reduced, and efficiency can be improved.

Description

technical field [0001] The invention relates to the technical field of voice services, in particular to a method and system for updating voice recognition resources. Background technique [0002] Currently, in the prior art, there is a lack of a complete solution for testing and updating speech recognition services. For services other than speech recognition, service updates can be carried out through offline testing, full or incremental update applications, and other methods. Due to its timeliness requirements, speech recognition services need to be updated frequently and stably, and the degree of automation should be high. [0003] Traditional service update technology usually updates the service in the test environment as a whole before going online, uses full or incremental methods to test service functions, and arranges to test key functions again after going online. The disadvantages of this update method are as follows: [0004] The cost of going online is high: a f...

AI Technical Summary

Problems solved by technology

[0004] The cost of going online is high: a full update of the service requires a lot of manpower, which will also affect the operation of the service

[0005] Affects system stability: The service update method is easy to encounter unexpected problems during the upgrade process, which is not suitable for speech recognition services: speech recognition services have strict requirements on timeliness and frequency, which cannot be met by traditional methods

[0006] In the process of realizing the present invention, the inventor found that the speech recognition resources and applications are completely decoupled. (including speech recognition resources and applications) are upgraded and updated, which leads to problems such as high cost of updates and upgrades, low efficiency, and error-prone.

Images