Spreading Activation for Resource Relevance Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current entity ranking methods fail to effectively combine unstructured and structured information within a single framework, lacking the ability to model communities of resources and utilize authorship information, which limits their accuracy in relevance rankings according to user expectations.
Innovation Solution
The system calculates relevance measures for computational resources by discovering latent topics, communities, and resource distributions using topic modeling and spreading activation over a structured graph, incorporating unstructured textual data and user-provided keywords to derive a final relevance ranking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If probabilistic models are used to measure associations between experts and resources, then relevance ranking can be performed, but the ability to model communities and utilize authorship information is limited
Solution Approach 1:
The patent merges probabilistic topic modeling with graph-based community detection to create a unified framework that simultaneously handles unstructured textual data and structured relational data. This combination enables the system to capture both local authorship patterns and global community structures, resolving the contradiction between measurement precision and adaptability.
Solution Approach 2:
The system implements a multi-functional approach where a single integrated model can process multiple types of information (textual content, authorship attributes, and relational graphs) and perform multiple tasks (topic discovery, community detection, and relevance ranking). This universality allows the system to maintain high accuracy while adapting to various data structures.
2Adaptability or versatility
If graph-based approaches are used to utilize predefined interconnections between entities, then structured information can be processed, but the ability to handle unstructured textual contents is limited
Solution Approach 1:
The patent combines graph-based community detection algorithms with probabilistic topic modeling to create a hybrid system that processes both structured relational data and unstructured textual data. The graph component captures predefined interconnections while the topic modeling component extracts semantic meaning from text, achieving both adaptability and precision.
Solution Approach 2:
The system introduces latent topics as intermediary representations that bridge the gap between structured graph data and unstructured text data. These latent topics serve as a common representation space where both types of information can be integrated and processed together, enabling the system to handle both structured and unstructured contents effectively.
3Measurement precision
If current topic modeling techniques are used, then probabilistic distributions over words can be estimated, but the ability to combine community modeling with authorship information is limited
Solution Approach 1:
The patent segments the complex modeling task into distinct but interconnected components: topic modeling for unstructured text, community detection for structured graphs, and authorship modeling for relational data. This segmentation allows each component to be optimized independently while maintaining overall coherence, managing complexity through modular design.
Solution Approach 2:
The system implements a nested hierarchical structure where topics are organized into communities, and communities are associated with authors. This nested arrangement allows the model to capture multiple levels of organization simultaneously, from fine-grained topic-level patterns to coarse-grained author-level characteristics, without proportionally increasing overall complexity.
Data Source
AI summary
The present invention relates to computer implemented methods and system for determining relevance measures for computational resources based on their relatedness to a user's interests. The methods and systems are designed to accept as inputs a collection of unstructured textual data related to resources, and a structured graph of the relationships between resources, to calculate probability distributions of resources over latent communities discovered from the unstructured textual data, to activate the structured graph with these probability distributions, and to spread this activation throughout the graph in a fixed number of iterations. The result of these methods and of the systems implementing these methods is a set of relevance measures attached to the resources in the structured graph.