US 12,015,737 B2
Methods, systems and apparatus for generating and/or using communications training data
Shaun Jaikarran Bharrat, Edison, NJ (US); and Michael Stovenour, McKinney, TX (US)
Assigned to Ribbon Communications Operating Company, Inc., Westford, MA (US)
Filed by Ribbon Communications Operating Company, Inc., Westford, MA (US)
Filed on Jun. 23, 2022, as Appl. No. 17/848,231.
Claims priority of provisional application 63/347,016, filed on May 30, 2022.
Prior Publication US 2023/0388414 A1, Nov. 30, 2023
Int. Cl. H04M 3/436 (2006.01); G06N 20/00 (2019.01)
CPC H04M 3/436 (2013.01) [G06N 20/00 (2019.01)] 25 Claims
OG exemplary drawing
 
1. A method of processing communications comprising:
generating, by a first network equipment device, from a first plurality of communications passing through a first communications network, a first set of communications media fingerprints and corresponding communications information, said communications media fingerprints, included in said first set of communications media fingerprints and corresponding communications information, being a first set of communications media fingerprints;
generating, by a second network equipment device, from a second plurality of communications, a set of communications media fingerprints of a first type, said set of communications media fingerprints of a first type being a second set of communications media fingerprints, each individual communication of said second plurality of communications having characteristics indicating a probability greater than a first threshold that the individual communication is of said first type;
labeling, by a training dataset generator, individual communications media fingerprints, in said first set of communications media fingerprints and corresponding communications information, said step of labeling including labeling communications media fingerprints, in said first set of communications media fingerprints and corresponding communications information, as being of said first type when the individual communications media fingerprint being labeled matches a communications media fingerprint in said set of communications media fingerprints of said first type; and
training, by a model generator, a model to identify communications of the first type, said training including using one or more labeled communications media fingerprints and corresponding communications information, from the first set of communications media fingerprints and corresponding communications information, to train the model to identify communications of the first type.