Visual blending of content

Patent metadata

Field	Value
Publication number	US-11416910-B1
Application number	US-202017034294-A
Country	US
Kind code	B1
Filing date	Sep 28, 2020
Priority date	Sep 28, 2020
Publication date	Aug 16, 2022
Grant date	Aug 16, 2022

Abstract

Systems and techniques are generally described for generating visually blended recommendation grids. In some examples, a selection of a first item and a second item displayed on a display may be received. In various examples, the first item may be displayed in a first element of a grid and the second item may be displayed in a second element of the grid. In some examples, a third element of the grid that is disposed between the first element and the second element along an axis of the grid may be determined. In various examples, a third item may be determined for display in the third element of the grid based at least in part on a blended representation of an embedding of the first item and an embedding of the second item. The third item may be displayed in the third element of the grid.

First claim

What is claimed is:

1. A method of displaying content, comprising:
displaying a first grid comprising a plurality of positions arranged in a first coordinate system;
displaying a first plurality of items, wherein each position of the first grid includes a respective item of the first plurality of items;
receiving a selection of a first item and a second item of the first plurality of items;
displaying the first item at a first corner position of the first grid;
displaying the second item at a second corner position of the first grid;
determining, for a non-corner position in the first grid, a first Euclidean distance between the non-corner position and the first corner position;
determining, for the non-corner position in the first grid, a second Euclidean distance between the non-corner position and the second corner position;
determining a first embedding vector representing the first item in an embedding space;
determining a second embedding vector representing the second item in the embedding space;
determining an inverse of the first Euclidean distance;
determining an inverse of the second Euclidean distance;
determining a first weighted value based on multiplying the inverse of the first Euclidean distance by the first embedding vector;
determining a second weighted value based on multiplying the inverse of the second Euclidean distance by the second embedding vector;
generating a weighted sum for the non-corner position in the first grid by adding the first weighted value and the second weighted value;
determining a closest embedding vector in the embedding space to the weighted sum;
determining a third item that is associated with the closest embedding vector; and
displaying the third item in the non-corner position of the first grid.

2. The method of claim 1 , further comprising:
receiving a selection of a graphical control position; and
displaying, in response to the selection of the graphical control position, an item detail page comprising:
a description of the first item;
one or more visual attributes associated with the first item;
a description of the second item; and
one or more visual attributes associated with the second item.

3. The method of claim 1 , further comprising:
receiving a selection of the third item;
receiving a selection of a graphical control configured to generate recommendations based on the third item;
displaying a second grid, wherein an image of the third item is displayed in a center position of the second grid;
determining, for an position of the second grid, a Euclidean distance from the center position of the second grid;
determining a set of items associated with the Euclidean distance;
sampling a fourth item from the set of items; and
displaying the fourth item in the position of the second grid.

4. A method comprising:
receiving a selection of a first item displayed on a display;
receiving a selection of a second item displayed on a display;
determining a first category associated with the first item and the second item;
determining, using a clustering algorithm, a number of clusters of items of the first category in an embedding space, wherein the number of clusters corresponds to a number of positions in a presentation;
generating first code, that when executed, is configured to cause the presentation to be displayed in a metric space, the presentation displaying the first item in a first position in the presentation and displaying the second item in a second position of the presentation, wherein each position of the presentation corresponds to a respective one of the clusters;
determining a third position of the presentation that is disposed between the first position and the second position in the metric space;
determining a third item for the third position of the presentation based at least in part on a blended representation of an embedding of the first item and an embedding of the second item; and
causing the third item to be displayed in the third position of the presentation.

5. The method of claim 4 , further comprising:
determining a first distance between the first position and the third position in the metric space; and
determining a second distance between the second position and the third position in the metric space, and wherein the blended representation is generated based at least in part on the first distance and the second distance.

6. The method of claim 5 , further comprising:
determining a first product of the embedding of the first item and an inverse of the first distance; and
determining a second product of the embedding of the second item and an inverse of the second distance.

7. The method of claim 6 , further comprising:
determining a sum of the first product and the second product, wherein the blended representation corresponds to the sum of the first product and the second product; and
determining a closest embedding in an embedding space to the sum of the first product and the second product.

8. The method of claim 7 , further comprising determining that the third item corresponds to the closest embedding in the embedding space.

9. The method of claim 4 , further comprising:
receiving a selection of a graphical control element; and
generating second code, in response to the selection of the graphical control element, the second code configured to cause an item detail page to be displayed, the item detail page comprising:
a description of the first item;
one or more visual attributes associated with the first item;
a description of the second item; and
one or more visual attributes associated with the second item.

10. The method of claim 4 , wherein items displayed along an axis of the metric space in the presentation represent a similarity continuum from a first end of the axis to a second end of the axis.

11. The method of claim 4 , wherein the presentation comprises a grid, and wherein the first position represents a first corner element of the grid and the second position represents a second corner element of the grid.

12. A method comprising:
determining a first category of items; and
determining, using a clustering algorithm, a number of clusters of items of the first category of items in an embedding space, wherein the number of clusters corresponds to a number of positions in a first presentation;
generating first code that, when executed, is configured to cause a first plurality of items arranged in the first presentation to be displayed, wherein each item of the first plurality of items corresponds to a respective one of the clusters;
receiving a selection of a first item from among the first plurality of items;
generating second code that, when executed, is configured to cause the first item to be displayed in a first position of a second presentation, wherein the second presentation represents a metric space;
determining, for a second position of the second presentation, a first distance between the first position and the second position in the metric space;
determining a first embedding vector for the first item;
determining a second item for the second position based at least in part on the first embedding vector and the first distance; and
generating third code that, when executed, is configured to cause the second item to be displayed in the second position of the second presentation.

13. The method of claim 12 , further comprising:
receiving a selection of a third item from among the first plurality of items in the first presentation; and
generating fourth code that, when executed, is configured to cause the third item to be displayed in a corner position of the secon

Assignees

• Amazon Tech Inc

Inventors

• Bazzani Loris
• Saina Filip
• Salvador Aguilera Amaia
• Martinez Gonzalez Angel Noe
• Vig Eleonora
• Gundogdu Erhan
• Donoser Michael

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/088

  Non-supervised learning, e.g. competitive learning · CPC title

• G06F3/04842

  Selection of displayed objects or displayed text elements (G06F3/0482 takes precedence) · CPC title

• G06F3/0483

  Interaction with page-structured environments, e.g. book metaphor · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title