Computer-implemented method for punctuation of text from audio input
Disclosed herein is a computer-implemented method for punctuation of text from audio. The method includes obtaining an audio input comprising speech data; identifying a plurality of silent sections in the audio input; grouping the plurality of silent sections into a plurality of groups, where each group in the plurality of groups corresponds to a punctuation mark or a space without a punctuation mark; and associating each silent section in the plurality of silent sections with a punctuation mark or a space according to the grouping of the silent sections, thus obtaining punctuation information.
1. A computer-implemented method for punctuation of text from an audio input, the method comprising:
obtaining an audio input comprising speech data;
identifying a plurality of silent sections in the audio input;
obtaining a type input indicating a type of the speech data in the audio input, wherein the type input indicates that the speech data is a customer service call, a public speech, or a lecture;
choosing an expected distribution, indicating an expected relative frequency of each group in a plurality of groups, according to the type input;
grouping the plurality of silent sections into the plurality of groups, wherein each group in the plurality of groups corresponds to a punctuation mark or a space without a punctuation mark, wherein the grouping the plurality of silent sections into the plurality of groups is done at least partially by using the expected distribution; and
associating each silent section in the plurality of silent sections with a punctuation mark or a space according to the group of the silent section, thus obtaining punctuation information,
performing a speech-to-text conversion on the audio input, thus obtaining a transcript of the speech data; and
punctuating the transcript according to the punctuation information by associating each silent section in the plurality of silent sections with the corresponding punctuation mark or the corresponding space without the punctuation mark;
wherein the expected distribution is based on statistical information about the plurality of silent sections between spoken words to apply punctuation.
2. The computer-implemented method according to claim 1 , wherein each group in the plurality of groups corresponds to a range of silent section temporal duration.
3. The computer-implemented method according to claim 1 , wherein the grouping the plurality of silent sections into the plurality of groups is done using a clustering algorithm.
4. The computer-implemented method according to claim 3 , wherein the clustering algorithm comprises k-means clustering.
5. The computer-implemented method according to claim 1 , wherein the expected distribution indicating an expected relative frequency of each group in the plurality of groups is at least partially based on an expected distribution of punctuation marks.
6. The computer-implemented method according to claim 5 , wherein the grouping the plurality of silent sections into the plurality of groups comprises:
determining at least one threshold temporal duration based on the expected distribution, wherein the at least one threshold temporal duration corresponds to a threshold between two groups in the plurality of groups; and
grouping the plurality of silent sections into the plurality of groups by comparing the temporal duration of each silent section in the plurality of silent sections to the at least one threshold temporal duration.
7. The computer-implemented method according to claim 5 , wherein a silent section in the plurality of silent sections is grouped into the plurality of groups using the expected distribution at least when the silent section cannot be grouped based on silent section temporal duration.
8. The computer-implemented method according to claim 1 , further comprising:
performing a speech-to-text conversion on the audio input, thus obtaining a transcript of the speech data; and
punctuating the transcript according to the punctuation information.
9. The computer-implemented method according to claim 8 , wherein the speech-to-text conversion further produces time stamp information of the plurality of silent sections, and wherein the plurality of silent sections is identified based on the time stamp information.
10. A computing device, comprising:
at least one processor; and
at least one memory including computer program code, the at least one memory and the computer program code configured to, with the at least one processor, cause the computing device to perform the method according to claim 1 .
11. A non-transitory computer program product comprising program code configured to perform the method according to claim 1 when the computer program product is executed on a computer.