Knowledge-Graph GNN Content Personalization for Buyer Journeys
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional ML models for predicting touchpoints in buyer decision journeys face challenges such as overfitting, underfitting, class imbalance, computational complexity, and difficulty in feature engineering, leading to suboptimal performance and failure to personalize content effectively, thus not guiding buyers efficiently through the decision journey.
Innovation Solution
The touchpoint content adaptation system uses a Graph Neural Network (GNN) model to encode buyer journey data into a knowledge graph, generating buyer embeddings that inform downstream classifiers to predict decision phases and touchpoints, and employs a large language model (LLM) to generate personalized multimedia content with specific tone and inflection to guide buyers to the next stage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional ML models are used for predicting touchpoints, then the system can process buyer journey data, but the prediction accuracy is suboptimal due to overfitting, underfitting, and class imbalance
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary structure that encodes domain knowledge about buyer journeys, touchpoints, and decision phases. This knowledge graph serves as a mediator between raw buyer journey data and the ML model predictions, providing contextual constraints and relationships that guide the model toward more accurate and reliable predictions while reducing overfitting and class imbalance issues
Solution Approach 2:
The patent transforms the ML modeling approach by changing from traditional feature-based inputs to knowledge graph-embedded representations. This parameter change involves encoding buyer journey data into graph structures with nodes representing buyers, touchpoints, and decision phases, and edges representing transitions, thereby improving prediction accuracy and model reliability
2Device complexity
If traditional ML models are used, then the system structure is simpler, but feature engineering is difficult and computational complexity increases
Solution Approach 1:
The knowledge graph structure enables self-service feature engineering where the graph embedding process automatically extracts relevant features and relationships from the encoded buyer journey data. The graph neural network performs automated feature extraction and transformation, eliminating the need for manual feature engineering while simplifying the overall system structure
Solution Approach 2:
The patent performs preliminary encoding of buyer journey data into a knowledge graph structure before ML model processing. This preliminary action organizes data into predefined schemas with nodes and edges representing buyers, touchpoints, and decision phases, making the subsequent ML processing more efficient and reducing computational complexity
3Productivity
If real-time content personalization is implemented, then buyer journey guidance is improved, but computational resources and processing time increase
Solution Approach 1:
The system performs preliminary encoding of buyer journey data into a knowledge graph structure and pre-computes graph embeddings for buyers, touchpoints, and decision phases. This preliminary action stores processed representations that can be quickly retrieved and used for real-time predictions, enabling content personalization without excessive processing time
Solution Approach 2:
The knowledge graph structure enables local quality processing where the system focuses computational resources on encoding and processing only the specific buyer journey data relevant to the current prediction task. The graph neural network processes local subgraphs and relationships pertinent to the buyer's current decision phase, reducing overall computational resource requirements while maintaining real-time performance
Data Source
AI summary
Techniques for personalizing multimedia content based on a knowledge graph are described. In one embodiment, a method includes receiving activity data associated with a user from a device, generating a touchpoint embedding and a decision embedding using a graph neural network (GNN) model based on the activity data, the GNN model trained using a knowledge graph, predicting a touchpoint using a first classifier based on the touchpoint embedding, predicting a decision stage using a second classifier based on the decision embedding, and generating personalized content for the touchpoint based on the decision stage using a large language model (LLM). Other embodiments are described and claimed.