Natural Language Content Embeddings for Accurate Ad Placement

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing natural language processing systems for digital ad campaigns are limited by content taxonomies, require manual effort and keyword accuracy, and struggle with user errors, leading to irrelevant ad placements.

Innovation Solution

A system using generative AI to create content embeddings for ad placements, allowing users to specify arbitrary descriptions and leveraging language models to capture in-context meanings, reducing reliance on predefined categories and keyword matching.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If content taxonomy is used for ad targeting, then ad placement structure is simplified, but ad placement accuracy deteriorates due to pre-defined category limitations

Engineering Contradiction:

Improvetargeting system structureVSAvoidad placement accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent replaces the mechanical taxonomy-based categorization system with an AI language model that processes natural language descriptions. Instead of forcing content into pre-defined categories, the system uses semantic understanding to match ads with content based on user intent, thereby maintaining simplicity while improving accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the parameter of content representation from fixed taxonomy categories to dynamic semantic embeddings generated by language models. This allows the same content to be represented differently based on contextual meaning, improving matching accuracy without requiring a complex hierarchical taxonomy structure.

Inventive Principle:

Principle #35Parameter changes

2Loss of time

If keyword matching is used for ad targeting, then manual effort is reduced, but ad placement accuracy deteriorates due to keyword ambiguity and lack of contextual understanding

Engineering Contradiction:

Improvemanual keyword selection timeVSAvoidad placement accuracy

Core Design Contradiction:

Loss of timeVSMeasurement precision

Solution Approach 1:

The patent replaces manual keyword selection and simple keyword matching algorithms with an AI language model that automatically understands user intent and generates appropriate content representations. This substitution eliminates the need for manual keyword research while providing contextually accurate ad placements.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The language model acts as an intermediary between the user's natural language description and the ad placement system. Instead of directly matching keywords, the model translates user intent into semantic representations that capture contextual meaning, thereby improving accuracy without requiring manual keyword optimization.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Device complexity

If keyword-based targeting is used, then system complexity is reduced, but robustness to user error deteriorates due to sensitivity to typos and extraneous phrases

Engineering Contradiction:

Improvetargeting system complexityVSAvoidrobustness to user error

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent implements error tolerance by designing the language model to understand and correct common user errors before they affect ad placement. The model's semantic understanding capability acts as a buffer that absorbs typos, extraneous phrases, and minor inaccuracies in user input, maintaining reliable performance even when users make mistakes.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

4Adaptability or versatility

If arbitrary natural language descriptions are accepted, then user flexibility is improved, but processing complexity increases due to need for semantic understanding

Engineering Contradiction:

Improveuser description flexibilityVSAvoidlanguage processing complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a universal language model that handles multiple functions: understanding user intent, generating semantic representations, and performing ad placement matching. This single multi-functional system accepts arbitrary natural language descriptions while managing processing complexity through the model's inherent language understanding capabilities.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20260017467A1Natural language processing of synthetic descriptions

Publication Date: 2026.01.15 VALASSIS DIGITAL CORP

AI summary

Content is identified based on natural language processing. A data object comprising a freeform text description of desired content is received. Supplementary data objects comprising other freeform text data representing of the desired content are created using a language model. The data object and the supplementary data objects are embedded into a vector space using a second language model. From a plurality of potential content objects, selected content objects are selected based on distances in the vector space between i) the selected content objects and ii) at least one of the data object and the supplementary data objects.