| CPC G06N 20/00 (2019.01) [G06F 16/334 (2019.01); G06F 16/583 (2019.01); G06F 16/783 (2019.01); G06F 16/9536 (2019.01); G06N 5/02 (2013.01); G06N 7/01 (2023.01)] | 20 Claims |
|
1. A method implemented by one or more processors, comprising:
for each electronic communication of a corpus of shared electronic communications each shared by a corresponding user to a platform and each accessible to one or more corresponding other users of the platform:
determining a set of features for the electronic communication, including at least:
a first feature that is based on content of the electronic communication, and
a second feature that is based on a quantity of past interactions with the electronic communication in a temporal period;
generating a request agnostic predicted interaction score for the electronic communication based on processing the set of features using a trained machine learning model; and
assigning, in one or more databases, the request agnostic predicted interaction score to the electronic communication;
subsequent to assigning the request agnostic predicted interaction scores to the electronic communications and subsequent to the electronic being shared to the platform:
detecting, based on monitoring each of the electronic communications, of the corpus of shared electronic communications, for occurrence of a corresponding rescoring event, occurrence of the corresponding rescoring event for a particular electronic communication of the corpus of shared electronic communications;
causing, in response to detecting occurrence of the corresponding rescoring event, the request agnostic predicted interaction score for the particular electronic communication of the corpus of shared electronic communications to be updated based on an additional quantity of past interactions with the particular electronic communication of the corpus of shared electronic communications subsequent to the temporal period;
receiving an electronic communications request for a client device;
identifying a plurality of the electronic communications of the corpus that are responsive to the electronic communications request;
retrieving, from the one or more databases, the request agnostic predicted interaction scores, including the request agnostic predicted interaction score that was updated based on the additional quantity of past interactions with the particular electronic communication of the corpus of shared electronic communications subsequent to the temporal period;
selecting, based at least in part on the request agnostic predicted interaction scores, including the request agnostic predicted interaction score that was updated based on the additional quantity of past interactions with the particular electronic communication of the corpus of shared electronic communications subsequent to the temporal period, a group of the identified plurality of the electronic communications to provide responsive to the request, wherein the request agnostic predicted interaction scores, used in the selecting, are generated independent of any content associated with the request and independent of any attributes associated with the request; and
transmitting, to the client device, content from each of the identified plurality of the electronic communications of the group.
|