Back to results
US 2020/0027463 A1Published application
Speaker diarization with cluster transfer
AssigneeInternational Business Machines Corporation
InventorsAaron K. BAUGHMAN, Stephen C. HAMMER
Life of this patent
Estimating the term…
Abstract
Methods, computer program products, and systems are presented. The methods include, for instance: obtaining the media file with a speech and identifying speakers on clusters separated by disfluencies and change of speakers. Clusters are re-segmented rearranged during diarization. Speaker identifications for the clusters in the media file is produced.
Claims20
A computer implemented method comprising:
- obtaining a media file having a speech by one or more speaker;
- stratifying the media file in order into a plurality of sessions at points of respective disfluencies in the speech;
- initializing one or more cluster space including one or more cluster, wherein the one or more cluster space respectively corresponds to the one or more speaker, wherein the one or more cluster results from identifying respective speaker of a predetermined number of session from the plurality of sessions;
- generating a cluster, wherein the generating a cluster includes identifying a speaker of a next session;
- determining a cluster action for the cluster based on a cluster epicenter;
- performing the cluster action from the determining on the one or more cluster space;
- bootstrapping the one or more cluster space resulting from the performing;
- generating another cluster by identifying speaker corresponding to another session from rest of the sessions;
- and
- producing speaker identifications respective to each cluster from the generating.
7 more elements
A computer program product comprising:
- a computer readable storage medium readable by one or more processing circuit and storing instructions for execution by one or more processor for performing a method comprising:
- obtaining a media file having a speech by one or more speaker;
- stratifying the media file in order into a plurality of sessions at points of respective disfluencies in the speech;
- initializing one or more cluster space including one or more cluster, wherein the one or more cluster space respectively corresponds to the one or more speaker, wherein the one or more cluster results from identifying respective speaker of a predetermined number of session from the plurality of sessions;
- generating a cluster, wherein the generating a cluster includes identifying a speaker of a next session;
- determining a cluster action for the cluster based on a cluster epicenter;
- performing the cluster action from the determining on the one or more cluster space;
- bootstrapping the one or more cluster space resulting from the performing;
- generating another cluster by identifying speaker corresponding to another session from rest of the sessions;
- and
- producing speaker identifications respective to each cluster from the generating.
8 more elements
A system comprising:
- a memory;
- one or more processor in communication with the memory;
- and
- program instructions executable by the one or more processor via the memory to perform a method comprising:
- obtaining a media file having a speech by one or more speaker;
- stratifying the media file in order into a plurality of sessions at points of respective disfluencies in the speech;
- initializing one or more cluster space including one or more cluster, wherein the one or more cluster space respectively corresponds to the one or more speaker, wherein the one or more cluster results from identifying respective speaker of a predetermined number of session from the plurality of sessions;
- generating a cluster, wherein the generating a cluster includes identifying a speaker of a next session;
- determining a cluster action for the cluster based on a cluster epicenter;
- performing the cluster action from the determining on the one or more cluster space;
- bootstrapping the one or more cluster space resulting from the performing;
- generating another cluster by identifying speaker corresponding to another session from rest of the sessions;
- and
- producing speaker identifications respective to each cluster from the generating.
11 more elements
Dependencies are inferred from the claim text.