Search Query Refinement via User Behavior Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search systems face inefficiencies due to slow bandwidth and resource bottlenecks when retrieving search results, leading to a suboptimal user experience, particularly in eCommerce platforms where users often conduct multiple actions to find items.
Innovation Solution
Implementing a system that processes search queries by receiving user inputs, analyzing previous search queries, and determining purchase probabilities to rank and display items, while clustering items and search terms based on user behavior, thereby refining search results and reducing system load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the system retrieves search results using traditional methods, then comprehensive search results can be provided, but system bandwidth becomes slow and bottlenecks occur
Solution Approach 1:
The system performs preliminary actions by analyzing user behavior patterns, clickstream data, and purchase history before the actual search query is executed. This pre-computation of user preferences and item relationships enables faster retrieval during the actual search operation, resolving the contradiction between comprehensive results and system speed.
Solution Approach 2:
The patent extracts and isolates the most relevant items and search terms based on user behavior analysis, separating them from the complete database. By extracting only the necessary subset of data for each user query, the system provides comprehensive results for the user's needs without processing the entire database, thus maintaining speed while ensuring reliability.
2Ease of operation
If users conduct numerous actions to find items, then they can locate desired products, but system resources are increased and efficiency decreases
Solution Approach 1:
The system implements continuous feedback loops by monitoring user interactions, clicks, and purchases, then using this feedback to refine and personalize search results in real-time. This feedback mechanism allows users to find items more easily through personalized recommendations while the system efficiently processes queries by learning from past behavior patterns, thereby maintaining both ease of operation and system productivity.
Solution Approach 2:
By pre-analyzing user behavior patterns and preparing personalized search profiles before actual search operations, the system reduces the number of actions users need to take. The preliminary preparation of user-specific search parameters and item rankings enables direct presentation of relevant results, improving ease of operation while reducing system resource consumption compared to processing multiple sequential user actions.
3Quantity of substance
If traditional search retrieval methods are used, then all items can be returned, but system resource demands increase
Solution Approach 1:
The patent applies local quality by providing different quantities and types of search results based on individual user characteristics, search context, and item relevance. Instead of uniformly returning all items to every user, the system tailors the search result set to each user's specific needs and behavior patterns, maintaining adequate result quantity while significantly reducing system resource consumption through targeted, personalized retrieval.
Data Source
AI summary
In many embodiments, a system comprising one or more processor and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform: determining a set of questions associated with a campaign by extracting text from one or more advertisements; generating vector embeddings of one or more online activities of a user; receiving a search query from a graphical user interface of a computing device of the user; determining a second question from the set of questions to present to the user based on respective relevance probabilities; determining a confidence score associated with the second question; and when the confidence score associated with the second question exceeds a second predetermined threshold, presenting the second question to the user via the graphical user interface. Other embodiments of related methods and systems are also provided.