Conference record generation method based on voice recognition, device and storage medium

Abstract

The invention discloses a conference record generation method based on voice recognition, and further discloses an electronic device and a computer storage medium. The conference record generation method comprises the steps that a conference record generation instruction sent by a user is received, and audio to be converted is acquired; sentence division is performed on the to-be-converted audio,and audio sentences of the to-be-converted audio are obtained; voiceprint features are extracted from the recognized audio sentences separately, the voiceprint features corresponding to the audio sentences are compared with a preset voiceprint feature library for analysis to determine speaker identity information corresponding to each audio sentence, and the audio sentences are divided into voicesegments according to the speaker identity information to determine a voice segment set corresponding to the to-be-converted audio; target voice recognition models corresponding to the voice segmentsare called to sequentially obtain text corresponding to each voice segment; and a conference record corresponding to the to-be-converted audio is generated. By using the conference record generation method based on voice recognition, accuracy and efficiency of conference record generation can be improved.

Description

technical field [0001] The present invention relates to the technical field of the Internet, in particular to a voice recognition-based conference record generating method, an electronic device, and a computer-readable storage medium. Background technique [0002] At present, the main methods of writing meeting minutes are as follows: first, key words are recorded at the meeting site; secondly, after the meeting, keywords are searched in the meeting recording and the recording near the key words is relistened to expand the key words to form a meeting record. However, since there is no corresponding relationship between keywords and recordings, recorders need to manually locate and search for keywords after the meeting, which is time-consuming and troublesome to operate. Furthermore, if the same keyword appears multiple times in the meeting , Only relying on manual re-listening to the recording positioning may cause wrong positioning, resulting in recording errors in the meet...

AI Technical Summary

Problems solved by technology

However, since there is no corresponding relationship between keywords and recordings, recorders need to manually locate and search for keywords after the meeting, which is time-consuming and troublesome to operate. Furthermore, if the same keyword appears multiple times in the meeting , only relying on manual re-listening of the recording positioning may lead to wrong positioning, resulting in recording errors in the meeting records

[0003] In order to solve the above-mentioned problems, there are conference record products that rely on voice conversion technology to automatically generate meeting record texts on the market. However, this existing conference record product is usually a simple voice-to-text product, and the accuracy of voice conversion is very high. Without guarantee, what the recorder gets after using it is a long text, which is not linked to the recording of the meeting. In addition, the voice-to-text technology is not mature enough, and the recorder often has no way to start after getting the text due to many transcribing errors. In the end, we can only rely on manual listening to the recording to complete the meeting minutes

Images