Semantic Neural Network for SKU Vector Representation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems face challenges in efficiently processing large volumes of data to recommend relevant products due to sparse data issues when modeling unique identification codes, such as stock keeping units (SKUs), leading to ineffective latent representations for interaction prediction and recommendation tasks.
Innovation Solution
A computer-implemented method using a semantic neural network model that generates a vector representation for unique identification codes through natural language processing, specifically employing algorithms like word2vec to learn informative representations by aggregating context codes and modifying weight matrices based on prediction errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data modeling is performed directly on unique identification codes, then the system can process identification codes, but the sparse data problem results in ineffective latent representations
Solution Approach 1:
The patent introduces context codes as an intermediary element between unique identification codes and latent representations. By modeling the contextual relationships surrounding each identification code (such as co-occurring items, user behaviors, or product attributes), the system enriches sparse identification code data with additional contextual information, enabling more effective latent representation learning without directly modeling the sparse identification codes alone
Solution Approach 2:
The patent transforms the problem from directly modeling unique identification codes in one dimension to modeling contextual relationships across multiple dimensions. By creating context vectors that capture relationships in additional dimensions (such as temporal, spatial, or categorical dimensions), the system overcomes the limitations of sparse direct modeling and generates richer latent representations
2Reliability
If large volumes of user activity data are processed to generate recommendations, then recommendation relevance improves, but processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary action by pre-computing context codes and latent representations during periods when user activity data is being collected. The system builds contextual models and generates vector representations in advance, so that when recommendations are needed, the system can quickly retrieve and use pre-computed representations rather than processing raw data in real-time, significantly reducing recommendation generation time
Solution Approach 2:
The patent replaces the mechanical process of real-time data processing with a computational approach using neural network models. By substituting traditional rule-based or statistical recommendation methods with learned vector representations and similarity computations, the system achieves both improved recommendation relevance through capturing complex patterns and reduced processing time through efficient vector operations
Data Source
AI summary
A computer-implemented method and system are described for learning a vector representation for unique identification codes. An example method may include generating a unique identification code list using one or more virtual interaction contexts, the unique identification code list being a list of unique identification codes, selecting a target unique identification code in the unique identification code list, and determining, from the unique identification code list, an input set of unique identification codes using the target unique identification code, the input set including the target unique identification code and one or more context unique identification codes. Some implementations may further include inputting the input set of unique identification codes into a semantic neural network model, the semantic neural network model including one or more weight matrices, and modifying the one or more weight matrices using the input set of unique identification codes.