Interest Graph Search Relevance via User Behavior Affinity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current enterprise search systems within organizations are inefficient, returning hundreds or thousands of irrelevant results due to the lack of user behavior data and popularity metrics, making it difficult for employees to find the information they need.
Innovation Solution
The creation and utilization of an interest graph that computes the affinity between users and information items based on user behavior, allowing for better ranking and retrieval of relevant information by analyzing user interactions and social connections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If text retrieval and content indexing are used for enterprise search, then documents can be identified and retrieved, but the search results are irrelevant and poorly ranked due to lack of user behavior data
Solution Approach 1:
The patent introduces an interest graph as an intermediary data structure that mediates between raw user behavior data and search results. The interest graph captures affinity relationships between users and information items, transforming raw interaction data into meaningful relevance rankings that resolve the contradiction between retrieving documents and presenting relevant results.
Solution Approach 2:
The patent changes the fundamental parameter for ranking from text-based content indexing to user-interest-based affinity scoring. By computing interest scores based on user behavior patterns and social connections rather than solely on keyword matching, the system achieves superior search relevance while utilizing available user behavior data.
2Productivity
If content indexing is relied upon for search, then documents can be found, but the system performs poorly because it lacks popularity, rating, or activity information
Solution Approach 1:
The patent performs preliminary computation of interest graphs during off-peak times, pre-calculating affinity scores between users and information items based on historical user behavior. This preliminary action enables the search system to quickly retrieve and rank relevant information without needing to process large quantities of user signal data in real-time during search operations.
3Ease of operation
If traditional search systems are used in enterprises, then basic text search is available, but employees give up on search systems and rely on alternative means to find information
Solution Approach 1:
The patent implements feedback mechanisms where user interactions with search results (such as viewing, downloading, or sharing) are captured and used to refine interest graph computations. This feedback loop continuously improves the accuracy of affinity scoring, making the search system more reliable and effective over time, thereby reducing employee frustration and dependency on alternative information finding methods.
Data Source
AI summary
A method, which identifies information of interest within an organization, determines use data that characterizes relationships among information items within the organization, where the information items include user data and collections of information items. The method generates interest data indicating affinities among the information items based on the determined use data. After receiving a query for data regarding the information items, the method responds to the query by providing one or more results based on the generated interest data. More details are provided herein.