Multipartite Graph Search Using Node Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search systems fail to accurately reflect user intent due to limited scope in processing search queries, leading to inefficient and inaccurate retrieval of data, which increases computational load and storage issues.
Innovation Solution
The use of a multipartite graph and node embedding algorithm to identify aspects based on past user behavior, generating vector representations that enable more accurate and efficient search results by learning relations between search queries, items, and their aspects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If search systems only consider current search query words for providing search results, then the search scope is limited and processing is simpler, but the accuracy of reflecting user intent deteriorates
Solution Approach 1:
The system pre-generates aspect vectors for all items in the database using historical data before search queries are submitted. This preliminary computation stores the relationships between items and their aspects in advance, allowing the search system to quickly compare query vectors against pre-computed aspect vectors without performing complex computations during actual search operations, thus improving accuracy while maintaining operational efficiency
Solution Approach 2:
The patent introduces aspect vectors as intermediary representations that bridge search queries and items. Instead of directly matching queries to items, the system converts both into aspect vectors and computes similarity in the vector space. This intermediary representation captures semantic relationships and user intent more effectively, improving search accuracy without proportionally increasing computational complexity
2Measurement precision
If search systems perform comprehensive processing to improve search accuracy, then user intent is better reflected, but computational load and storage demands increase
Solution Approach 1:
Aspect vectors are pre-computed and stored for all items using historical interaction data before search operations. This shifts the computational burden to an offline preprocessing stage, allowing online search operations to simply compare query vectors against pre-computed aspect vectors using efficient similarity metrics like cosine similarity, dramatically reducing online computational load while maintaining high search accuracy
Solution Approach 2:
The system creates vector representations (copies) of item aspects that capture the essential semantic information. Instead of processing original complex item data and historical interactions during each search, the system uses these pre-generated vector copies for rapid comparison and matching, reducing computational load while preserving search accuracy
3Productivity
If search systems use simple query matching, then processing is faster and more efficient, but the scope and accuracy of search results are limited
Solution Approach 1:
The patent replaces traditional mechanical text-matching mechanisms with vector-based similarity computation. Instead of keyword matching algorithms, the system converts queries and items into vectors and uses geometric similarity measures (like cosine similarity) to determine relevance. This substitution enables the system to capture semantic meaning and user intent more effectively while maintaining computational efficiency through optimized vector operations
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The technology disclosed herein relates to identifying an aspect from a search query based on using a multipartite graph generated using past user behavior and a node embedding algorithm for determining vector representations of nodes of the multipartite graph. For example, nodes of the multipartite graph can include nodes for prior search queries, items or item listings associated with the prior search queries, and one or more of an aspect or category of the items or item listings. In embodiments, the multipartite graph has dynamic edges between the nodes for the prior search queries and the items or item listings. In embodiments, a query expansion is performed based on identifying the aspect using the multipartite graph and node embedding algorithm. In embodiments, search results are provided based on identifying the aspect and performing the query expansion. For example, one or more identified aspects can be provided as selectable options.