Personalized Digital Design Template Recommendation via Transformer Embeddings

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional digital design template recommendation systems are inefficient and inflexible, often requiring significant user interaction to find relevant templates, providing static and generic recommendations that fail to match individual creative intents, especially on mobile devices with limited screen space and numerous template options.

Innovation Solution

A personalized digital design template recommendation system generates embedding vectors using a transformer to understand template metadata and user activity, utilizing a similarity search model to provide real-time, contextually relevant recommendations by comparing user embedding vectors with template vectors, dynamically updating based on individual user events and geo-seasonal intents.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If conventional recommendation systems provide generic template recommendations, then system complexity is reduced, but recommendation relevance to individual user creative intent deteriorates

Engineering Contradiction:

Improvesystem complexityVSAvoidrecommendation relevance

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The system changes the parameter of user representation from generic demographics to dynamic embedding vectors that capture individual creative intent. By continuously updating user embeddings based on real-time interactions with digital content, the system achieves personalized recommendations without requiring complex rule-based systems, resolving the contradiction between simplicity and relevance.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces mechanical recommendation systems (rule-based filtering) with machine learning-based embedding models. The transformer-based embeddings automatically learn and represent user preferences and template characteristics, substituting complex mechanical recommendation logic with data-driven semantic understanding that achieves higher relevance.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Manufacturing precision

If the system provides comprehensive template options, then template selection quality is improved, but user navigation time and device screen space requirements increase

Engineering Contradiction:

Improvetemplate selection qualityVSAvoiduser navigation time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The system performs preliminary action by pre-computing embedding vectors for all templates and organizing them in a searchable vector database before user interaction. This allows the system to instantly retrieve relevant templates through similarity search without requiring users to navigate through comprehensive option lists, maintaining high selection quality while eliminating navigation time loss.

Inventive Principle:

Principle #10Preliminary action

3Adaptability or versatility

If the system updates recommendations in real-time based on user activity, then recommendation adaptability is improved, but computational resource consumption increases

Engineering Contradiction:

Improverecommendation adaptabilityVSAvoidcomputational resource consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The system applies partial action by selectively updating only the user embedding vector based on specific interaction types (e.g., likes, shares, completion events) rather than continuously processing all user activities. This approach maintains recommendation adaptability by responding to significant user actions while avoiding excessive computational resource consumption from processing every minor interaction.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20240338406A1Recommendating personalized digital design templates utilizing content embeddings

Publication Date: 2024.10.10 ADOBE INC

AI summary

The present disclosure relates to systems, methods, and non-transitory computer-readable media that utilizes embedding vectors to identify a subset of digital design templates to recommend to a client device based on individual user events. For instance, the disclosed systems extract metadata from digital design templates and generate, utilizing a transformer, a plurality of embedding vectors for the digital design templates from the extracted metadata. Further, the disclosed system generates, utilizing the transformer, a user embedding vector from one or more user events of a user. Moreover, the disclosed system utilizes identifies, utilizing a similarity search model, a subset of digital design templates from the digital design templates to recommend to the user by identifying one or more embedding vectors of the plurality of embedding vectors that satisfy a similarity threshold to the user embedding vector.