Zero-Shot Image Moderation in Google Ads with LLM-Assisted Textual Descriptions and Cross-modal Co-embeddings

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Luo, Enming, Qiao, Wei, Warren, Katie, Li, Jingxiang, Xiao, Eric, Viswanathan, Krishna, Wang, Yuan, Liu, Yintao, Li, Jimin, Fuxman, Ariel
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912165142724608
author Luo, Enming
Qiao, Wei
Warren, Katie
Li, Jingxiang
Xiao, Eric
Viswanathan, Krishna
Wang, Yuan
Liu, Yintao
Li, Jimin
Fuxman, Ariel
author_facet Luo, Enming
Qiao, Wei
Warren, Katie
Li, Jingxiang
Xiao, Eric
Viswanathan, Krishna
Wang, Yuan
Liu, Yintao
Li, Jimin
Fuxman, Ariel
contents We present a scalable and agile approach for ads image content moderation at Google, addressing the challenges of moderating massive volumes of ads with diverse content and evolving policies. The proposed method utilizes human-curated textual descriptions and cross-modal text-image co-embeddings to enable zero-shot classification of policy violating ads images, bypassing the need for extensive supervised training data and human labeling. By leveraging large language models (LLMs) and user expertise, the system generates and refines a comprehensive set of textual descriptions representing policy guidelines. During inference, co-embedding similarity between incoming images and the textual descriptions serves as a reliable signal for policy violation detection, enabling efficient and adaptable ads content moderation. Evaluation results demonstrate the efficacy of this framework in significantly boosting the detection of policy violating content.
format Preprint
id arxiv_https___arxiv_org_abs_2412_16215
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Zero-Shot Image Moderation in Google Ads with LLM-Assisted Textual Descriptions and Cross-modal Co-embeddings
Luo, Enming
Qiao, Wei
Warren, Katie
Li, Jingxiang
Xiao, Eric
Viswanathan, Krishna
Wang, Yuan
Liu, Yintao
Li, Jimin
Fuxman, Ariel
Computer Vision and Pattern Recognition
Artificial Intelligence
Information Retrieval
We present a scalable and agile approach for ads image content moderation at Google, addressing the challenges of moderating massive volumes of ads with diverse content and evolving policies. The proposed method utilizes human-curated textual descriptions and cross-modal text-image co-embeddings to enable zero-shot classification of policy violating ads images, bypassing the need for extensive supervised training data and human labeling. By leveraging large language models (LLMs) and user expertise, the system generates and refines a comprehensive set of textual descriptions representing policy guidelines. During inference, co-embedding similarity between incoming images and the textual descriptions serves as a reliable signal for policy violation detection, enabling efficient and adaptable ads content moderation. Evaluation results demonstrate the efficacy of this framework in significantly boosting the detection of policy violating content.
title Zero-Shot Image Moderation in Google Ads with LLM-Assisted Textual Descriptions and Cross-modal Co-embeddings
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Information Retrieval
url https://arxiv.org/abs/2412.16215