Context-based automatic speech recognition processing

A computer-implemented method comprising:

• receiving audio embedding data corresponding to a first portion of audio data;
• determining the first portion of the audio data is likely to include an entity;
• determining entity embedding data corresponding to at least a first entity;
• based on determining the first portion of the audio data is likely to include the entity, determining, using the entity embedding data and the audio embedding data, a biasing value corresponding to the at least first entity;
• determining weighted entity embedding data using the entity embedding data and the biasing value;
• determining, using the first portion of the audio data and the weighted entity embedding data, a first portion of automatic speech recognition (ASR) data;
• and
• determining, using the ASR data, output data responsive to the audio data.

A system comprising:

• at least one processor;
• and
• at least one memory including instructions that, when executed by the at least one processor, cause the system to:
• receive audio embedding data corresponding to a first portion of audio data;
• determine the first portion of the audio data is likely to include an entity;
• determine entity embedding data corresponding to at least a first entity;
• based on determining the first portion of the audio data is likely to include an entity, determine, using the entity embedding data and the audio embedding data, a biasing value corresponding to the at least first entity;
• determine weighted entity embedding data using the entity embedding data and the biasing value;
• determine, using the first portion of the audio data and the weighted entity embedding data, automatic speech recognition (ASR) data;
• and
• determine, using the ASR data, output data.