Extended Search Retrieval Using Graph-Based Intent Associations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated shopping assistants fail to capture all relevant products due to limitations in natural language processing, missing similar intents such as 'Automotive Batteries' when searching for 'Jeep Liberty Battery', leading to incomplete search results.
Innovation Solution
Implement a system that associates first and second intents based on similarity, using a bipartite graph and graph embedding methods to group similar intents, and extends the search query to include these associated intents, thereby enhancing the search retrieval process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If natural language processing is used to detect intent from user queries, then the interface becomes more intuitive and easier to use, but similar intents are not captured leading to incomplete search results
Solution Approach 1:
The system performs preliminary action by pre-establishing intent associations through graph embedding methods before actual search queries are received. The bipartite graph is constructed and processed in advance to create pre-computed associations between similar intents (e.g., Automotive Batteries and Vehicle Batteries), so that when a user queries for one intent, the associated intents are already prepared and can be immediately retrieved without additional real-time computation.
Solution Approach 2:
The patent introduces an intermediary mechanism - the bipartite graph with graph embedding - that mediates between the user's simple natural language query and the complex database of product intents. This intermediary layer translates the user's direct intent query into an expanded set of related intents by leveraging pre-computed graph relationships, thereby bridging the gap between intuitive user input and comprehensive search results.
2Loss of information
If the search query is extended to include similar intents, then the completeness of search results is improved, but the complexity of the search system increases
Solution Approach 1:
The system performs preliminary action by pre-establishing intent associations through graph embedding methods before actual search queries are received. The bipartite graph is constructed and processed in advance to create pre-computed associations between similar intents (e.g., Automotive Batteries and Vehicle Batteries), so that when a user queries for one intent, the associated intents are already prepared and can be immediately retrieved without additional real-time computation.
Solution Approach 2:
The patent applies copying by creating a simplified representation of the complex intent relationships through graph embedding. Instead of storing all possible intent combinations and relationships in full detail, the system creates a compressed graph structure that captures the essential relationships between similar intents. This embedded representation can be quickly queried and expanded during actual search operations, reducing the effective complexity while maintaining comprehensive coverage.
Data Source
AI summary
The disclosed subject matter relates to a system and method for providing an extended search. The system generates a list of similar intent groups based on previous engagements linking queries and intents. Upon the detection of an intent in a user query, the system accesses similar intents from the similar intent group and incorporates them into the query of the product catalog in order to obtain a complete set of results. The creation of the similar intent groups uses bipartite graphs and graph embedding to identify intents close in distance.