Content Feed Policy Enforcement via Adjacent Item Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing content feed management systems struggle to ensure that certain content items, such as those containing nudity, violence, or political content, are not placed adjacent to target content items due to lack of effective classification and filtering mechanisms.
Innovation Solution
A machine learning model is trained to classify content items into prohibited categories based on features and user engagement, and a content feed manager uses this model to determine whether to place a target content item between adjacent items by ensuring they are not in prohibited classifications, adhering to policies set by content providers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If audience-based topic exclusion control is used to filter content feeds, then user preference alignment is improved, but content provider policy compliance deteriorates
Solution Approach 1:
The patent segments the filtering mechanism into two distinct layers: (1) audience-based topic exclusion control that filters content based on user preferences and historical behaviors, and (2) a new policy-based filtering layer that prevents target content items from being placed adjacent to prohibited content items (nudity, violence, political content). This segmentation allows both user preference alignment and content provider policy compliance to be satisfied simultaneously by operating at different levels of the content feed management system.
2Productivity
If content items are placed adjacent to each other based on user behavior, then user engagement is improved, but harmful content placement deteriorates
Solution Approach 1:
The patent introduces a machine learning-based classification model as an intermediary between the content placement decision and the actual placement. This intermediary classifies content items into prohibited categories (nudity, violence, political content) and prevents target content items from being placed adjacent to classified prohibited content, even when user behavior suggests such placement would be engaging. This intermediary layer resolves the contradiction by filtering out harmful placements while preserving beneficial user engagement patterns.
3Reliability
If machine learning classification is applied to all adjacent content items, then policy compliance is improved, but system complexity deteriorates
Solution Approach 1:
The patent applies preliminary action by training the machine learning classification model in advance on a dataset of content items with labels indicating whether they belong to prohibited categories. The model is pre-trained to recognize patterns and features associated with nudity, violence, and political content. During content feed generation, the pre-trained model quickly classifies adjacent content items without requiring complex real-time analysis, thus maintaining policy compliance while reducing runtime system complexity.
Data Source
AI summary
A method for generating a content feed includes receiving, from a content provider, a target content item and a policy for the target content item, specifying a prohibited classification of the content that cannot be published adjacent to the target content item. The method also includes identifying a slot in a content feed including multiple content items. A machine learning model is then accessed and applied to two adjacent content items that are adjacent to the slot to determine whether each of the adjacent content items is the prohibited classification. Responsive to determining that the adjacent content items are not the prohibited classification, the target content item is placed in the slot, and the content feed including the target content item is sent for display to a viewing user.