Automatic speech recognition fusion system
AssigneeSnap Inc.
InventorsSharon Moll, Isac Andreas Müller Sandvik, Daniel Harris, Neken Aritia Symphonie Razafindrabe
Life of this patent
Estimating the term…
Abstract
An automatic fusion system analyzes received audio, using a first machine learning model, to generate a first textual transcription of the audio, provides the first textual transcription of the audio to a computing device, analyzes the received audio, using a second machine learning model, to generate a second textual transcription of the audio, and fuses the first textual transcription and the second textual transcription using a third machine learning model to generate a fused transcription.
Claims20
A system comprising:
- at least one processor;
- at least one memory component storing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
- receiving audio, via a computing device, and performing in real-time until termination is detected, operations comprising:
- analyzing the received audio, using a first machine learning model, to generate a first textual transcription of the audio;
- providing the first textual transcription of the audio to the computing device;
- analyzing the received audio, using a second machine learning model, to generate a second textual transcription of the audio;
- and
- fusing the first textual transcription and the second textual transcription using a third machine learning model to generate a fused transcription;
- and
- upon detection of termination of the audio, generating a final transcription by fusing the first textual transcription and the second textual transcription of all audio received, using the third machine learning model.
A computer-implemented method comprising:
- receiving audio, via a computing device, and performing in real-time until termination is detected, operations comprising:
- analyzing the received audio, using a first machine learning model, to generate a first textual transcription of the audio;
- providing the first textual transcription of the audio to the computing device;
- analyzing the received audio, using a second machine learning model, to generate a second textual transcription of the audio;
- and
- fusing the first textual transcription and the second textual transcription using a third machine learning model to generate a fused transcription;
- and
- upon detection of termination of the audio, generating a final transcription by fusing the first textual transcription and the second textual transcription of all audio received, using the third machine learning model.
A non-transitory computer-readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:
- receiving audio, via a computing device, and performing in real-time until termination is detected, operations comprising:
- analyzing the received audio, using a first machine learning model, to generate a first textual transcription of the audio;
- providing the first textual transcription of the audio to the computing device;
- analyzing the received audio, using a second machine learning model, to generate a second textual transcription of the audio;
- and
- fusing the first textual transcription and the second textual transcription using a third machine learning model to generate a fused transcription;
- and
- upon detection of termination of the audio, generating a final transcription by fusing the first textual transcription and the second textual transcription of all audio received, using the third machine learning model.
Dependencies are inferred from the claim text.