Punctuation generation method, apparatus, and storage medium

By splitting the pronunciation information of audio signals into fine-grained characters and using a model to predict punctuation mark indication information, the problem of low efficiency in punctuation mark generation in existing technologies is solved, and more efficient and accurate punctuation mark generation is achieved.

CN115775555BActive Publication Date: 2026-07-03HUAWEI TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HUAWEI TECH CO LTD
Filing Date
2021-09-06
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing automatic speech recognition technology suffers from large model space consumption, high energy consumption, and low efficiency when generating punctuation marks, making it difficult to efficiently and accurately generate punctuation marks in speech recognition results.

Method used

By splitting the pronunciation information of the audio signal into finer-grained characters, the first model is used to predict punctuation mark indication information, and combined with text information to determine the second text. This reduces the computational load of the model and improves the accuracy and efficiency of punctuation mark prediction by utilizing contextual information.

Benefits of technology

This technology enables more accurate generation of punctuation marks in speech recognition results, reduces the model's space usage and energy consumption, and improves the accuracy and efficiency of punctuation mark prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115775555B_ABST
    Figure CN115775555B_ABST
Patent Text Reader

Abstract

The application relates to the technical field of artificial intelligence, in particular to a punctuation generation method and device and a storage medium, the method comprising the following steps: obtaining pronunciation information corresponding to an audio signal; inputting at least one first character into a first model to obtain first punctuation symbol indication information, the first character being a character obtained by splitting the pronunciation information corresponding to the audio signal, and the first punctuation symbol indication information indicating a punctuation symbol corresponding to each first character; determining a second text according to the first character, the first punctuation symbol indication information and a first text, wherein the first text comprises a character corresponding to the audio signal, and the second text comprises the character corresponding to the audio signal and punctuation symbols. According to the embodiment of the application, the model occupied space and energy consumption can be reduced, and a more efficient method for generating punctuation in a speech recognition result is realized.
Need to check novelty before this filing date? Find Prior Art