Federated Search Ranking via Relative Order Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Federated search engines face challenges in merging and ranking search results from disparate resources due to distinct ranking heuristics used by each resource, leading to inconsistencies and potential errors in presenting the most relevant results to users.
Innovation Solution
A method and system that determine a unified ranking for search results from multiple resources by preserving the relative ranking of each resource, using metadata such as title, summary, and relevancy scores, without requiring access to full documents, thus maintaining the local ranking of each resource and reducing computational errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the federated search engine combines relevancy scores for results returned by multiple resources, then the results can be reranked based on combined scores, but this may lead to purely interleaved results where some results are given an unduly amount of importance over others
Solution Approach 1:
The patent segments the ranking process by preserving the relative ordering within each resource's result set as a separate entity. Instead of completely mixing all results into a single pool, it maintains distinct ranked lists from each resource and interleaves them while preserving internal rankings. This segmentation prevents any single resource from dominating the overall ranking while maintaining the reliability of each resource's own ranking judgment.
Solution Approach 2:
The patent applies local quality by allowing each search resource to maintain its own ranking quality standards and heuristics independently. Each resource's results are ranked according to its own criteria, and this local ranking quality is preserved in the unified results. The federated search engine respects the local ranking decisions of each resource rather than imposing a uniform ranking approach that might compromise individual resource quality.
2Measurement precision
If the federated search engine re-ranks results based on a consistent relevancy calculation requiring full text access, then relevancy can be calculated accurately, but this requires downloading each search result and may be error prone
Solution Approach 1:
The patent extracts only the necessary metadata fields from each search result rather than downloading and processing full text documents. By taking out only the essential metadata components needed for ranking (such as titles, abstracts, and other structured fields), the system achieves accurate relevancy calculation without the computational burden of full text processing, thereby reducing device complexity while maintaining measurement precision.
Solution Approach 2:
The patent applies partial action by performing relevancy calculation on a subset of result attributes (metadata) rather than complete document processing. This partial processing approach provides sufficient ranking accuracy for federated search purposes without the excessive computational resources required for full text analysis, effectively balancing measurement precision with reduced computational complexity.
3Reliability
If the federated search engine creates a profile for each federated resource and submits queries only to authoritative resources, then resource authority can be optimized, but this may lead to a longer response time for presenting results
Solution Approach 1:
The patent performs preliminary action by pre-establishing resource profiles and authority assessments before actual search queries are executed. By determining resource authority levels in advance and storing this information in profiles, the system can quickly route queries to appropriate resources without performing time-consuming authority evaluations during the actual search process, thus maintaining reliability while reducing response time.
Solution Approach 2:
The patent applies dynamics by making the resource selection process adaptive rather than static. The system dynamically adjusts which resources are queried based on the specific query characteristics and pre-established authority profiles. This dynamic approach allows the system to optimize for both resource authority and response time by selecting the most appropriate subset of resources for each query rather than following a fixed, time-consuming protocol.
Data Source
AI summary
Techniques are disclosed for generating a unified ranking of search results received from a plurality of distinct search resources. A computing system executing a federated search application receives a set of ranked search results from multiple search resources (e.g., web-based search engines, public databases and catalogs, etc.). Each search result identifies an item and includes metadata characterizing the item. The federated search application determines a unified ranking for the items identified in the search results that preserves a relative ranking of each set of items, such that a result for a given item is ranked above other results for items in the associated set of items is ranked above the other results in the unified ranking.