Citation-Based Document Ranking Algorithm
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search tools rely heavily on term frequency and similarity calculations using word vectors, which may not adequately capture the relevance of documents, particularly in complex queries, as they ignore important factors beyond term frequency and context.
Innovation Solution
The method calculates a normalized activity score for ranking documents by considering citation networks, metadata, and community recognition within a subject matter community, incorporating IR scores weighted by citations and community prominence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional search tools use term frequency and word vector similarity to rank documents, then the ranking process is simple and fast, but the relevance accuracy is insufficient particularly for complex queries
Solution Approach 1:
The patent combines multiple ranking factors including term frequency, word vector similarity, citation counts, and community recognition metrics into a unified ranking system. This merging of multiple information sources and evaluation criteria enables more accurate relevance assessment for complex queries while maintaining a structured approach to complexity management.
Solution Approach 2:
The ranking system is designed to handle both simple and complex queries effectively by incorporating universal metrics that work across different query types. The system uses multi-functional evaluation criteria that can adapt to various document types and search contexts, making it universally applicable while improving accuracy.
2Adaptability or versatility
If search tools rank documents based on term frequency alone, then the ranking method is simple, but it ignores important factors beyond term frequency and context
Solution Approach 1:
The patent introduces additional dimensions to the ranking process beyond traditional term frequency. It incorporates citation network analysis, community recognition metrics, and contextual relevance factors as new dimensions. This multi-dimensional approach enables comprehensive evaluation of document relevance while systematically managing the increased complexity through structured integration of multiple factors.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computerized method is provided for calculating a normalized activity score value to rank an identified document. The method comprises identifying a stored document; determining a number of times the identified document was cited in a subject matter community of the identified document; determining a probability distribution that individual documents within the subject matter community are cited a variable number of times by other individual documents in the subject matter community; calculating a probability function by performing a regression on the probability distribution; calculating the activity score value according to an activity score function formulated as an inverse of the probability function; weighting the activity score value by an age of the identified document; and storing in computer memory a ranking of the identified document based on the activity score value.