Transcript alignment
View Patent ↗An approach to alignment of transcripts with recorded audio is tolerant of moderate transcript inaccuracies, untranscribed speech, and significant non-speech noise. In one aspect, a number of search terms are formed from the transcript such that each search term is associated with a location within the transcript. Possible locations of the search terms are then determined in the audio recording. The audio recording and the transcript are then aligned using the possible locations of the search terms. In another aspect a search expression is accepted, and then a search is performed for spoken occurrences of the search expression in an audio recording. This search includes searching for text occurrences of the search expression in a text transcript of the audio recording, and searching for spoken occurrences of the search expression in the audio recording.
1. A method comprising:
accepting a search expression;
searching for text occurrences of the search expression in a transcript of an audio recording;
searching for spoken occurrences of the search expression in the audio recording without requiring first searching for the text occurrences of the search expression in the transcript of the audio recording;
compiling results of the searching for spoken occurrences and text occurrences, wherein the compiling includes eliminating duplicate occurrences of a common instance of the search expression according to time alignments of text occurrences and spoken occurrences; and
presenting representations of the compiled results of the searching, including enabling access to portions of the audio recording corresponding to speech occurrences and text occurrences in the results of the searching.
2. The method of claim 1 wherein the searching for the text occurrences and the searching for the spoken occurrences are performed concurrently.
3. The method of claim 1 wherein presenting the representations of the results includes sorting the results according to confidence that the search expression is present in the corresponding portions of the audio recording.
4. The method of claim 1 further comprising:
prior to accepting the search expression, storing an alignment of the text transcript of the audio recording with the audio recording, said alignment associating at least some portions of the text transcript with corresponding portions of the audio recording.
5. The method of claim 4 wherein enabling access to the portions of the audio recording corresponding to each of the results includes accessing the stored alignment of the text transcript to identify the portions of the audio recording corresponding to at least some of the results.
6. The method of claim 1 further comprising:
prior to accepting the search expression, preprocessing the audio recording to determine statistical quantities related to presence of spoken units in the audio recording.
7. The method of claim 6 wherein searching for the spoken occurrences of the search expression includes accessing the statistical quantities determined during preprocessing of the stored audio recording.
8. The method of claim 1 wherein searching for the spoken occurrences in the audio recording includes applying a word spotting approach.
9. The method of claim 1 further comprising:
accepting a selection of one of the results of the searching; and
playing the portion of the audio recording corresponding to the selected result.
10. The method of claim 9 further comprising:
presenting a graphical representation of a portion of the text transcript corresponding to the selected result.
11. The method of claim 1 , wherein searching for the text occurrences comprises searching for the text occurrences of the search expression in a time-aligned transcript of the audio recording.
12. The method of claim 11 , wherein compiling the results of the searching comprises using time offsets provided in the time-aligned transcript to eliminate duplicate occurrences of a common instance of the search expression.
13. One or more processor readable storage devices having code embodied on said storage devices, said code for programming one or more processors to perform a method comprising:
accepting a search expression;
searching for text occurrences of the search expression in a transcript of an audio recording;
searching for spoken occurrences of the search expression in the audio recording without requiring first searching for the text occurrences of the search expression in the transcript of the audio recording;
compiling results of the searching for spoken occurrences and text occurrences, wherein the compiling includes eliminating duplicate occurrences of a common instance of the search expression according to time alignments of text occurrences and spoken occurrences; and
presenting representations of the compiled results of the searching, including enabling access to portions of the audio recording corresponding to speech occurrences and text occurrences in the results of the searching.
14. The storage devices of claim 13 wherein the storage devices further have embodied thereon:
data encoding the audio recording;
the text transcript of the audio recording; and
data associating portions of the text transcript with corresponding portions of the audio recording.
15. The storage devices of claim 13 wherein the one or more storage devices include one or more removable optical disks.