公开(公告)号：US20240282107A1

公开(公告)日：2024-08-22

申请号：US18452424

申请日：2023-08-18

Applicant: Apple Inc.

Inventor： Jia Huang ,  Robert J. Monarch ,  Alex Jungho Kim ,  Jungsuk Kwac ,  Parmeshwar Khurd ,  Kailash Thiyagarajan ,  Xiaoyuan Goodman Gu

IPC: G06V20/30 ,  G06F16/53 ,  G06V10/774

CPC classification number: G06V20/30 ,  G06F16/53 ,  G06V10/774 ,  G06V10/82

Abstract: The present technology pertains to a multi-modal transformer model that is designed and trained to perform cross-modal tasks such as image-text matching, wherein the model is further refined with data for the particular downstream use case of the model. More specifically, the present technology can refine the underlying model with labeled examples derived from a dataset of text-image pairs that ultimately achieved a desired interaction in the proper context. For example, in the use case of advertising applications in an App store, the present technology can refine the underlying model with examples of images used to advertise applications in the App store where the respective invitational content was clicked or converted.

公开(公告)号：US20240281490A1

公开(公告)日：2024-08-22

申请号：US18458817

申请日：2023-08-30

Applicant: Apple Inc.

Inventor： Xiaoyuan Goodman Gu ,  Yuan Yen Tai ,  Moliang Zhou ,  Dayvid Victor Rodrigues de Oliveira ,  Brian Knott ,  Jin Cao ,  Jia Huang

IPC: G06F16/954 ,  G06F16/951

CPC classification number: G06F16/954 ,  G06F16/951

Abstract: The present technology has the ability to establish connections between content that do not have direct or explicit relationships. Implicit influence relationships can be established from user download sequence data, campaign data with keyword targeting, and content review data that mentions other content. Using these influence relationships, the relevance of content items can be determined based on the influence relationship of linked content items and a similarity relationship of content items. However, the importance of the influence relationship in ranking content items can vary depending on the parameters against which the content item is considered relevant. To address this, the present technology includes a context-driven factor that is used as a weight to adjust the impact of the influence relationship of the ranking, depending on the parameters against which the content item is considered relevant.