User preference guided content generation from paired comparisons

1 . A system for guiding generation of items comprises:

a computing device comprising one or more processors;

a neural network;

a transceiver; and

at least one memory in communication with the computing device, the neural network, and the transceiver, the at least one memory storing computer program code that, when executed by the computing device, is configured to cause the system to:

output a first set of items having a first attribute with a first feedback prompt to a user device;

receive from a user a first user input of a selection of a selected item in the first set of items;

store the first user input to the neural network;

determine, based on the first user input, a user preference for the first attribute, the user preference free of any textual articulation for the first attribute;

generate, using the neural network, a second set of items comprising a first item and one or more additional items, the first item of the second set comprising a first configuration of a first modified first attribute, the one or more additional items of the second set comprising one or more additional configurations of the first modified first attribute, wherein the first modified first attribute is based, at least in part, on the user preference for the first attribute, and wherein the first configuration and the one or more additional configurations of the first modified first attribute are configured to be different;

output the second set of items having the first modified first attribute with a second feedback prompt to a user device;

receive from the user a second user input of a selection of a selected item in the second set of items;

store the second user input to the neural network;

determine, based on at least the second user input, an updated user preference for the first attribute, the updated user preference free of any textual articulation for the first attribute; and

generate, using the neural network, a third set of items comprising a second modified first attribute, wherein the second modified first attribute is based, at least in part, on the updated user preference for the first attribute.

2 . The system of claim 1 , wherein the neural network is configured to map similar user preferences related to the first modified first attribute and a second modified second attribute closely together within a structured latent space to improve the system prediction and generation of at least one of the additional items.

3 . The system of claim 1 , wherein:

the second user input is received from the user device; and

the second user input comprises information related to the selection of the selected item in the second set of items from among the first item and the one or more additional items.

4 . The system of claim 3 , wherein the computer program code, when executed by the computing device, is further configured to cause the system to:

output the third set of items having the second modified first attribute with a third feedback prompt to the user device;

wherein:

the third set of items comprises a first item and one or more additional items;

the first item of the third set comprises a first configuration of the second modified first attribute;

the one or more additional items of the third set comprise one or more additional configurations of second modified first attribute; and

the first configuration and the one or more additional configurations of the second modified first attribute are configured to be different.

5 . The system of claim 4 , wherein the computer program code, when executed by the computing device, is further configured to cause the system to:

store feedback data on the neural network;

predict a user preference to an adjustment of at least one of:

the first modified first attribute of the first item of the second set or one or more of the additional items of the second set of items; or

the second modified first attribute of the first item of the third set or one or more of the additional items of the third set; and

generate one or more new items based, at least in part, on the prediction of the user preference related to the adjustment of at least one of the first or second modified first attribute.

6 . The system of claim 5 , wherein the computer program code, when executed by the computing device, is further configured to cause the system to:

adjust and generate one or more of the new items until the adjustment of at least one of the first or second modified first attribute meets a desired preference.

7 . A method for guiding generation of images comprising:

outputting a first set of images having a first attribute with a first feedback prompt to a user device;

receiving from the user device a first user input of a selection of a selected image in the first set of images;

storing the first user input to a neural network;

determining, based on the first user input, a user preference for the first attribute;

generating, using the neural network, a second set of images comprising a first image and one or more additional images, the first image of the second set comprising a first configuration of a first modified first attribute, the one or more additional images of the second set comprising one or more additional configurations of the first modified first attribute, wherein the first modified first attribute is based, at least in part, on the user preference for the first attribute, and wherein the first configuration and the one or more additional configurations of the first modified first attribute are configured to be different;

outputting the second set of images having the first modified first attribute with a second feedback prompt to the user device;

receiving from the user device a second user input of a selection of a selected image in the second set of images;

determining, based on the first and second user inputs, an updated user preference for the first attribute;

generating, using the neural network, a third set of images comprising a second modified first attribute, wherein the second modified first attribute is based, at least in part, on the updated user preference for the first attribute; and

mapping, with the neural network, similar user preferences related to the first modified first attribute and a second modified second attribute closely together within a generative model latent space to improve the system prediction and generation of at least one additional image.

8 . The method of claim 7 , wherein the generative model latent space is a structured latent space derived from an encoder or Contrastive Language-Image Pre-training (CLIP) embeddings.

9 . The method of claim 7 , wherein the second user input comprises information related to the selection of the selected image in the second set of images from among the first image and the one or more additional images.

10 . The method of claim 9 further comprising:

outputting the third set of images with a third feedback prompt to the user device;

wherein:

the third set of images comprises a first image and one or more additional images;

the first image of the third set comprises a first configuration of the second modified first attribute;

the one or more additional images of the third set comprise one or more additional configurations of second modified first attribute; and

the first configuration and the one or more additional configurations of the second modified first attribute are configured to be different.

11 . The method of claim 10 further comprising:

storing feedback data on the neural network;

predicting a user preference to an adjustment of at least one of:

the first modified first attribute of the first image of the second set or one or more of the additional images of the second set;

the second modified first attribute of the first image of the third set or one or more of the additional images of the third set; and

generating one or more new images based, at least in part, on the prediction of the user preference related to the adjustment of at least one of the first or second modified first attribute.

12 . The method of claim 11 further comprising:

adjusting and generating one or more of the new images until the adjustment of at least one of the first or second modified first attribute meets a desired preference.

13 . A non-transitory computer readable medium having stored thereon instructions comprising executable code for guiding generation of items, when executed by one or more processors, causes one or more of the processors to:

output a first set of items having a first attribute with a first feedback prompt to a user device;

receive from the user device a first user input of a preference of a selection of a selected item in the first set of items, wherein the selection does not require a user to articulate the preference textually;

store the first user input to a neural network;

determine, based on the first user input, a user preference for the first attribute;

generate, using the neural network, a second set of items comprising a first item and one or more additional items, the first item of the second set comprising a first configuration of a first modified first attribute, the one or more additional items of the second set comprising one or more additional configurations of the first modified first attribute, wherein the first modified first attribute is based, at least in part, on the user preference for the first attribute, and wherein the first configuration and the one or more additional configurations of the first modified first attribute are configured to be different;

output the second set of items having the first modified first attribute with a second feedback prompt to a user device;

receive from the user device a second user input of a preference of a selection of a selected item in the second set of items, wherein the selection does not require the user to articulate the preference textually;

determine, based on the first and second user inputs, an updated user preference for the first attribute; and

generate, using the neural network, a third set of items comprising a second modified first attribute, wherein the second modified first attribute is based, at least in part, on the updated user preference for the first attribute.

14 . The non-transitory computer readable medium of claim 13 , wherein the neural network is configured to map similar user preferences related to the first modified first attribute and a second modified second attribute closely together within a structured latent space derived from an encoder or Contrastive Language-Image Pre-training (CLIP) embeddings to improve a system prediction and generation of at least one of the additional items.

15 . The non-transitory computer readable medium of claim 13 , the second user input comprises information related to the selection of the selected item in the second set of items from among the first item and the one or more additional items.

16 . The non-transitory computer readable medium of claim 15 , wherein the instructions, when executed by one or more of the processors, further cause one or more of the processors to:

output the third set of items with a third feedback prompt to the user device;

wherein:

the third set of items comprises a first item and one or more additional items;

the first item of the third set comprises a first configuration of the second modified first attribute;

the one or more additional items of the third set comprise one or more additional configurations of second modified first attribute; and

the first configuration and the one or more additional configurations of the second modified first attribute are configured to be different.

17 . The non-transitory computer readable medium of claim 16 , wherein the instructions, when executed by one or more of the processors, further cause one or more of the processors to:

store feedback data on the neural network;

predict a user preference to an adjustment of at least one of:

the first modified first attribute of the first item of the second set or one or more of the additional items of the second set; or

the second modified first attribute of the first item of the third set or one or more of the additional items of the third set;

generate one or more new items, based at least in part, on the prediction of the user preference related to the adjustment of at least one of the first or second modified first attribute; and

adjust and generate one or more of the new items until the adjustment of at least one of the first or second modified first attribute meets a desired preference.

18 . A system for guiding generation of items comprises:

a computing device comprising one or more processors;

a neural network;

a transceiver; and

at least one memory in communication with the computing device, the neural network, and the transceiver and storing computer program code that, when executed by the computing device, is configured to cause the system to:

output a first set of items having a first attribute;

receive from a user a first user input of a selection of a selected item in the first set of items;

store the first user input to the neural network;

generate, using the neural network, a second set of items comprising a first modified first attribute, wherein the first modified first attribute is based, at least in part, on a user preference for the first attribute;

output the second set of items having the first modified first attribute with a second feedback prompt to a user device;

receive from the user a second user input of a selection of a selected item in the second set of items;

determine, based on at least the second user input, an updated user preference for the first attribute; and

generate, using the neural network, a third set of items comprising a second modified first attribute, wherein the second modified first attribute is based, at least in part, on the updated user preference for the first attribute;

wherein:

the second set of items comprises a first item and one or more additional items;

the first item comprises a first configuration of the first modified first attribute;

the one or more additional items comprise one or more additional configurations of the first modified first attribute; and

the first configuration and the one or more additional configurations of the first modified first attribute are configured to be different.

19 . The system of claim 18 , wherein:

the second user input is received from the user device; and

the second user input comprises information related to the selection of the selected item in the second set of items from among the first item and the one or more additional items.

20 . The system of claim 19 , wherein the at least one memory further comprises computer program code that, when executed by the computing device, is further configured to cause the system to:

output the third set of items having the second modified first attribute with a third feedback prompt to the user device;

wherein:

the third set of items comprises a first item and one or more additional items;

the first item comprises a first configuration of the second modified first attribute;

the one or more additional items comprise one or more additional configurations of second modified first attribute; and

the first configuration and the one or more additional configurations of the second modified first attribute are configured to be different.

21 . The system of claim 20 , wherein the at least one memory further comprises computer program code that, when executed by the computing device, is further configured to cause the system to:

store feedback data on the neural network;

predict a user preference to an adjustment of at least one modified first attribute of the first item or one or more additional items of a set of items; and

generate one or more new items based, at least in part, on the prediction of the user preference related to the adjustment of the at least one modified first attribute.

22 . The system of claim 21 , wherein the at least one memory further comprises computer program code that, when executed by the computing device, is further configured to cause the system to:

adjust and generate one or more of the new items until the adjustment of the at least one modified first attribute meets a desired preference.