US 11,921,775 B2
Media unit retrieval and related processes
Michael Elkaim, London (GB); Michael Kopp, London (GB); and Kristjan Korjus, London (GB)
Assigned to DREAM IT GET IT LIMITED, London (GB)
Filed by DREAM IT GET IT LIMITED, London (GB)
Filed on Dec. 9, 2022, as Appl. No. 18/078,789.
Application 18/078,789 is a continuation of application No. 16/985,159, filed on Aug. 4, 2020, granted, now 11,567,989.
Application 16/985,159 is a continuation of application No. 15/756,502, granted, now 10,769,197, issued on Sep. 8, 2020, previously published as PCT/EP2016/070493, filed on Aug. 31, 2016.
Claims priority of application No. 15183394 (EP), filed on Sep. 1, 2015; application No. 15183395 (EP), filed on Sep. 1, 2015; application No. 15183396 (EP), filed on Sep. 1, 2015; application No. 15183397 (EP), filed on Sep. 1, 2015; application No. 15183398 (EP), filed on Sep. 1, 2015; application No. 15183399 (EP), filed on Sep. 1, 2015; application No. 15183401 (EP), filed on Sep. 1, 2015; and application No. 15183402 (EP), filed on Sep. 1, 2015.
Prior Publication US 2023/0124063 A1, Apr. 20, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 16/58 (2019.01); G06F 16/21 (2019.01); G06F 16/907 (2019.01); G06F 16/908 (2019.01); G06F 17/16 (2006.01); G06N 3/04 (2023.01)
CPC G06F 16/58 (2019.01) [G06F 16/211 (2019.01); G06F 16/907 (2019.01); G06F 16/908 (2019.01); G06F 17/16 (2013.01); G06N 3/04 (2013.01)] 12 Claims
OG exemplary drawing
 
1. A method of retrieving one or more media units or their respective media unit identifiers, each media unit being associated with one or more images with a respective feature set defining a set of one or more attribute values, the method comprising:
deriving one or more derived attribute values of the set of attribute values by transforming the one or more images associated with the media units, wherein the transforming comprises inputting the one or more media units into one or more neural networks, and deriving the one or more derived attribute values directly or indirectly from activations of neurons of a layer of the one or more neural networks;
receiving an input following presentation of a set of media units of the one or more media units;
updating an intent model over the one or more attribute values comprising the one or more derived attribute values, wherein the intent model comprises a probabilistic model and/or a geometric model,
the intent model defining a measure of intent with respect to each feature set, wherein the updating comprises using the input to determine which attribute values to use, to generate an updated intent model, wherein the updated intent model comprises a probabilistic model and/or a geometric model;
selecting a next set of media units of the one or more media units using the updated intent model; and
transmitting one or more media units or their respective media unit identifiers for presentation of the next set of one or more media units,
wherein the selecting of the next set of one or more media units or their media unit identifiers comprises using a selection function, and
wherein the selection function comprises computing a measure of similarity between (a) the set of attribute values of the one or more selected and/or not selected media units of the presented media units, and (b) the set of attribute values of one or more media units for potential inclusion in the next set of one or more media units or their respective media unit identifiers existing for presentation, and
wherein the measure of similarity is a distance or combination of distances between the attributes.