US 12,561,367 B1Granted

Enhanced machine learning-based micro-category generation

AssigneeAmazon Technologies, Inc.

InventorsCaizhi Xu, Ravi Shankar, Brian William Beisel, Gabriel Miller, Cody Severinski, Jamal Alikhani, Jose A. Sepulveda, Brian Robert Swickline, Fei Wang, Hani Suleiman

Life of this patent

Estimating the term…

patentgrep

Loading patent…

Abstract

Devices, systems, and methods for machine learning-based micro-category generation may include a method including identifying, for a respective cluster of content titles or items, a top-K most representative group of the content titles or items; generating, by a machine learning model, based on a first input prompt, the common theme of the top-K most representative group; generating, by the machine learning model, based on the second input prompt and filtering, a group of content titles or items matching the common theme; generating, by the machine learning model, based on a third input prompt, a name for the group of content titles or items matching the common theme; and presenting the group of content titles or items matching the common theme and the name for the group of content titles or items matching the common theme via a user interface of a streaming media application.

Claims20

Dependent claims

A method for machine-learning based micro-category generation, the method comprising:

identifying, for a respective cluster of video titles, a subset of the cluster, the subset consisting of a top-K most representative group of video titles in the cluster;
providing, to a large language model (LLM), a first input prompt comprising the top-K most representative group, a first task of returning a common theme of the subset, and title cookies for the subset, wherein the first input prompt excludes at least one of the video titles in the cluster and not in the subset;
generating, by the LLM, based on the first input prompt, the common theme of the subset;
providing, to the LLM, a second input prompt comprising the common theme, the title cookies, and a second task of filtering out any respective video title of the subset that does not match the common theme;
generating, by the LLM, based on the second input prompt and the filtering, a group of video titles in the subset and matching the common theme;
providing, to the LLM, a third input prompt comprising the group of video titles matching the common theme, the common theme, and a third task to generate a name for the group of video titles based on the common theme;
generating, by the LLM, based on the third input prompt, a name for the group of video titles matching the common theme;
and
presenting the cluster of video titles and the name for the group of video titles matching the common theme via a user interface of a streaming video application.

6 more elements

A method for machine-learning based micro-category generation, the method comprising:

providing, to a generative artificial intelligence language model, a first input prompt comprising a most representative subset of a group of content titles or items sharing characteristics, a first task of returning a common theme of the subset, and textual information for the subset, wherein the first input prompt excludes at least one of the content titles or items in the group and not in the most representative subset;
generating, by the generative artificial intelligence language model, based on the first input prompt, the common theme of the subset;
and
causing presentation of the group and the common theme via a user interface of a device.

1 more element

A system for machine-learning based micro-category generation, the system comprising:

at least one first processor associated with a generative artificial intelligence language model;
and
memory coupled to at least one second processor, wherein the at least one second processor is configured to:
provide, to the generative artificial intelligence language model, a first input prompt comprising a most representative subset of a group of content titles or items sharing characteristics, a first task of returning a common theme of the subset, and textual information for the subset wherein the first input prompt excludes at least one of the content titles or items in the group and not in the most representative subset;
generate, using the generative artificial intelligence language model, based on the first input prompt, the common theme of the subset;
and
cause presentation of the group and the common theme via a user interface of a device.

4 more elements

Dependencies are inferred from the claim text.