Graph-Based Query Refinement Clustering by User Intent
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search engines struggle to effectively cluster query refinements to represent diverse user information needs, often relying on frequency rather than diversity, which limits the selection and placement of relevant query suggestions on search result pages.
Innovation Solution
A graph-based clustering method that combines content similarity and session co-occurrence information, using a Markov model to identify transitions between queries and documents, allowing for the grouping of query refinements into clusters that reflect distinct user intents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If query refinements are selected based on frequency from query logs, then the most popular queries are displayed, but the diversity of user information needs is not adequately represented
Solution Approach 1:
The patent segments the set of query refinements into multiple clusters, where each cluster represents a distinct user intent or information need. Instead of displaying refinements as a flat list based on frequency, the system organizes them into thematic groups (e.g., geographic locations, business types, services), ensuring diverse coverage of user needs while maintaining relevance within each cluster.
2Adaptability or versatility
If many related queries are mined from query logs, then comprehensive coverage of user interests is achieved, but the limited space on the search results page cannot accommodate all queries
Solution Approach 1:
By clustering refinements into thematic groups, the patent enables efficient use of limited page space. Each cluster can be displayed as a compact unit with a representative query or category label, allowing comprehensive coverage of diverse user interests without requiring individual space for each refinement query.
Solution Approach 2:
The patent merges multiple related refinement queries into single clusters representing unified user intents. This consolidation allows the system to present comprehensive information coverage in a space-efficient manner, grouping semantically related queries together and displaying them as unified suggestions rather than individual entries.
3Quantity of substance
If query refinements are displayed without clustering, then all available refinements can be shown, but user understanding of diverse information needs is reduced and information overload occurs
Solution Approach 1:
The patent segments the large set of query refinements into organized clusters with meaningful groupings. This segmentation helps users understand the diversity of information needs by presenting refinements in structured categories rather than as an overwhelming unorganized list, improving ease of navigation and comprehension.
Solution Approach 2:
The patent introduces cluster representatives or category labels as intermediaries between the user and the full set of refinement queries. These intermediaries provide semantic context and organization, helping users understand the nature of each group of refinements without requiring them to examine every individual query option.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for clustering query refinements. One method includes building a representation of a graph for a first query, wherein the graph has a node for the first query, a node for each of a plurality of refinements for the first query, and a node for each document in the document sets of the refinements, and wherein the graph has edges from the first query node to each of the refinement nodes, edges from the first query to each document in the respective document set of the first query, edges from each refinement to each document in the respective document set of the refinement, and edges from each refinement to each co-occurring query of the refinement. The method further includes clustering the refinements into refinement clusters by partitioning the refinement nodes in the graph into proper subsets.