Audience Interest Estimation via Bipartite Graph Inference

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for predicting audience interest on web pages are limited as they either focus on explicit content-based targeting, fail to model general audience interests, or require costly and error-prone manual categorization, especially for complex or image-based websites.

Innovation Solution

A probabilistic model that estimates audience interest distribution based on users' anonymous web behavior, using a bipartite graph to infer user interest distributions and calculate expected audience interest for websites, without relying on user identities or content specifics.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual categorization is used to predict audience interest, then measurement precision may be improved, but device complexity and loss of time increase significantly

Engineering Contradiction:

Improveaudience interest prediction accuracyVSAvoidmanual categorization complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces manual categorization (mechanical human effort) with an automated probabilistic model that uses matrix factorization and bipartite graphs to infer audience interests from browsing behavior data. This substitution eliminates the need for manual content analysis while maintaining or improving prediction accuracy through scalable computational methods.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service by automatically inferring audience interests through probabilistic modeling of browsing behavior without requiring external manual intervention. The model autonomously processes browsing data, constructs user interest profiles, and generates audience interest predictions for web pages, freeing operators from manual categorization tasks.

Inventive Principle:

Principle #25Self-service

2Ease of operation

If content-based targeting is used, then ease of operation is improved, but adaptability to general audience interests deteriorates

Engineering Contradiction:

Improvetargeting implementation simplicityVSAvoidgeneral audience interest modeling

Core Design Contradiction:

Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent segments audience interest into multiple latent factors through matrix factorization, allowing the system to capture diverse and general audience interests beyond single content categories. This segmentation enables the model to represent complex user preferences as combinations of underlying interest dimensions, improving adaptability while maintaining operational simplicity through automated factor extraction.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The probabilistic model serves multiple functions: it infers individual user interests, predicts audience interest for any web page, and adapts to different browsing behavior patterns without requiring content-specific configuration. This universal approach handles both content-based and behavior-based targeting scenarios within a single framework.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Adaptability or versatility

If behavioral targeting is used to model user interests, then adaptability to audience interests is improved, but device complexity and loss of information increase

Engineering Contradiction:

Improveuser interest modeling capabilityVSAvoiduser identity privacy

Core Design Contradiction:

Adaptability or versatilityVSLoss of information

Solution Approach 1:

The patent introduces latent interest factors as intermediaries between raw browsing behavior data and user interest profiles. Instead of directly using identifiable user data, the model processes behavior through latent factor representations that capture interest patterns while obscuring individual user identities. This intermediary layer preserves adaptability to user interests while protecting privacy and reducing information loss.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Ease of operation

If contextual targeting is used, then ease of operation is improved, but adaptability to general audience interests deteriorates

Engineering Contradiction:

Improveadvertising placement simplicityVSAvoidaudience interest representation

Core Design Contradiction:

Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent adds a latent factor dimension to traditional contextual targeting by representing audience interests in a multi-dimensional latent space rather than relying solely on explicit content categories. This dimensional transformation allows the system to capture implicit and general audience interests that extend beyond the surface-level content context, improving versatility while maintaining operational simplicity through automated factor inference.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Data Source

PatentUS10599981B2System and method for estimating audience interest

Publication Date: 2020.03.24 NEW YORK UNIV

AI summary

Systems, methods and computer-accessible mediums can be provided that can determine an audience interest distribution(s) of content(s) by, for example, receiving first information related to a web behavior(s) of a user(s), determining second information related to a user interest distribution(s) of the user(s) based on the first information, and determining determine the audience interest distribution(s) of the content(s) based on the second information.