AI Publisher Identification via Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying the publisher associated with a content item is challenging, especially when only the identity of the writer and content item are available, due to the complexity of mappings between writers, content items, and publishers, which is crucial for licensing and royalty purposes.
Innovation Solution
A graph network is used to link entities like content item metadata and publishing metadata, with machine learning models trained to predict the most probable publisher associated with a given writer or content item, employing graph theory and neural networks to determine node distances and probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to identify publishers, then the process requires manual effort and complex mappings, but the identification accuracy and efficiency deteriorate
Solution Approach 1:
The patent introduces machine learning models as an intermediary between writer identities and publisher identities. These models are trained on historical data to learn the complex mapping relationships, thereby automating the identification process and improving accuracy without requiring manual intervention in the actual publisher identification task.
Solution Approach 2:
The patent replaces manual mechanical processes of publisher identification with automated machine learning systems. The ML models process writer identities and content item metadata to automatically predict publisher identities, eliminating the need for manual mapping and significantly improving efficiency and consistency.
2Productivity
If manual methods are used to determine publisher identities, then the process is simple to implement, but the time consumption and productivity deteriorate
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on extensive historical data containing writer identities, content items, and publisher identities. This training phase occurs beforehand, enabling the models to quickly and accurately predict publisher identities during actual operations without requiring real-time complex processing or manual intervention.
3Reliability
If complete publisher identification is attempted for all content items, then comprehensive coverage is achieved, but the computational resources and system complexity increase
Solution Approach 1:
The patent implements partial action by focusing the machine learning models on predicting the most probable publisher identities based on available input data (writer identity and content item metadata). Rather than attempting to exhaustively verify all possible publishers, the system predicts the most likely matches with high confidence, achieving practical completeness without excessive computational overhead.
Data Source
AI summary
In one embodiment, a computer-implemented method includes training, using an artificial intelligence engine, one or more machine learning models using training data comprising identities of writers of content items as input and to output a dataset comprising identities of publishers and a respective probability that each identity of a publisher, from the identities of publishers, is associated with a respective identity of a writer from the identities of writers; receiving, via the one or more machine learning models, a first identity of a first writer; inputting the first identity of the first writer into the one or more machine learning models; outputting, via the one or more machine learning models, the dataset comprising the identities of publishers and the respective probability that each identity of the publisher, from the identities of publishers, is associated with the respective identity of the writer from the identities of writers.