Hold back and real time ranking of results in a streaming matching system

A system, comprising:

• a receiving component configured to receive, from an audio streaming service, a probe audio sample;
• a comparing component configured to compare a first portion of the probe audio sample to reference audio samples;
• a matching component configured to identify sufficiently matching reference audio samples based on a first comparison between feature vectors of the first portion of the probe audio sample and respective first feature vectors of the reference audio samples;
• a retaining component configured to retain the sufficiently matching reference audio samples based on a determination that none of the sufficiently matching reference audio samples meet a predetermined score threshold,
• the comparing component is further configured to:
• identify additional matching reference audio samples based a second comparison of feature vectors of a second portion of the probe audio sample to second feature vectors of the reference audio samples;
• a ranking component configured to:
• assign first respective ranking scores to the sufficiently matching reference audio samples based on the first comparison;
• and
• assign second respective ranking scores to the additional matching reference audio samples based on the second comparison;
• and
• a scoring component configured to:
• update the sufficiently matching reference audio samples using the first respective ranking scores and second respective ranking scores;
• and
• output at least one of the reference audio samples based on weighting of the first respective ranking scores and the second respective ranking scores.

A non-transitory computer readable storage medium comprising computer executable instructions that, in response to execution, cause a computing system comprising a processor to perform operations, comprising:

• receiving a probe sample;
• comparing a first portion of the probe sample to reference samples to identify sufficiently matching reference samples based on a first comparison of feature vectors of the first portion of the probe sample and respective first feature vectors of the reference samples;
• assigning first respective ranking scores to the sufficiently matching reference audio samples based on a first comparison;
• and
• in response to determining that the sufficiently matching reference samples do not meet a predetermined score threshold:
• retaining the sufficiently matching reference samples;
• and
• comparing a second portion of the probe sample to the reference samples to identify additional matching reference samples based on a second comparison between feature vectors of the second portion of the probe sample and second feature vectors of the reference samples;
• assigning second respective ranking scores to additional matching reference audio samples;
• and
• outputting at least one of the reference samples based on weighting of the first respective ranking scores and the second respective ranking scores.