Citation Graph Centrality

Measures the raw number of direct connections a node has within the citation graph. A source with high degree centrality is cited by or cites many other documents directly.

• In-degree: Number of incoming citations from other sources
• Out-degree: Number of references a source makes to others
• High in-degree signals popularity and recognition within a field
• Example: A landmark paper cited by 5,000+ subsequent studies has extreme in-degree centrality

Quantifies how often a node serves as a bridge or gatekeeper between otherwise disconnected clusters in the citation network.

• Calculated by counting the fraction of shortest paths between all node pairs that pass through the target node
• High betweenness indicates a source that connects disparate research domains
• Signals interdisciplinary influence and knowledge synthesis capability
• Example: A review paper that links neuroscience findings to machine learning architectures

Measures how quickly information from a source can spread through the entire citation network based on its average distance to all other nodes.

• Computed as the inverse of the sum of shortest path distances to every other node
• High closeness means a source is topologically near the center of the graph
• Indicates rapid influence propagation across the research community
• Nodes with high closeness can disseminate findings faster than peripheral sources

Extends degree centrality by weighting connections based on the importance of the citing sources themselves. A citation from a highly authoritative source carries more weight.

• Based on the principle that connections to high-scoring nodes contribute more to authority
• Computed iteratively using the adjacency matrix's principal eigenvector
• Google's original PageRank algorithm is a variant of eigenvector centrality
• Example: A citation from Nature or Science confers more authority than one from an obscure journal

A random-walk based centrality metric originally designed for web graphs, adapted to measure citation authority with a damping factor that prevents rank sinks.

• Models a researcher randomly following citations, with occasional jumps to random papers
• The damping factor (typically 0.85) represents the probability of following a citation vs. teleporting
• Converges to a stable distribution representing stationary authority scores
• Handles directed graphs and mitigates manipulation better than simple citation counts

Identifies nodes that occupy gaps between dense clusters in the citation graph, providing unique brokerage advantages and access to non-redundant information.

• Measured by constraint and effective size metrics
• Low constraint indicates a node bridges otherwise disconnected communities
• Sources in structural holes have access to diverse, non-overlapping knowledge pools
• Correlates with innovation potential and novel idea recombination
• Example: A researcher whose work is cited independently by both theoretical physics and applied engineering communities

The foundational graph algorithm originally developed by Larry Page and Sergey Brin that assigns a numerical weighting to each node in a hyperlinked set of documents. PageRank operates on the principle that a link from a high-authority page is a more significant endorsement than a link from an obscure source. In the context of generative engines, this concept has been extended beyond web links to include citation edges in academic graphs, patent networks, and knowledge bases. The algorithm iteratively distributes rank through the graph until convergence, effectively identifying nodes that are not just popular, but are referenced by other authoritative nodes.

A measure of a node's influence based on how often it lies on the shortest path between other pairs of nodes in the citation graph. A source with high betweenness centrality acts as a critical bridge connecting disparate research clusters or knowledge domains. For AI-driven search, these bridging sources are disproportionately valuable because they serve as interdisciplinary validators, linking a claim in one field to supporting evidence in another. This metric identifies the gatekeepers of information flow, not just the most-cited nodes.

A recursive measure of influence where a node's score is proportional to the sum of the scores of its neighbors. Unlike simple citation counts, eigenvector centrality recognizes that a citation from a highly authoritative paper is worth exponentially more than a citation from a peripheral source. This metric is the mathematical backbone of the 'prestige' model in citation analysis. In generative engine optimization, content referenced by nodes with high eigenvector centrality is treated as high-confidence source material for factual grounding and direct answer generation.

Hyperlink-Induced Topic Search (HITS) is a link analysis algorithm that assigns two scores to each node: authority and hub. An authority is a node cited by many high-quality hubs, while a hub is a node that points to many high-quality authorities. This mutual reinforcement model creates a self-consistent ecosystem where good hubs and good authorities validate each other. For generative engines, distinguishing between hub-like content (curated lists, review papers) and authority-like content (original research, definitive sources) is critical for source provenance scoring.

A bibliometric technique that measures the relatedness of two documents based on how frequently they are cited together by a third document. Co-citation strength indicates semantic and topical proximity, forming the basis for clustering research fronts and identifying the intellectual structure of a field. In the context of generative engines, strong co-citation signals between a new piece of content and established authoritative sources can accelerate entity relationship novelty recognition, effectively bootstrapping a new document's authority by association with the existing citation core.

The ratio of actual citation edges to the total possible edges in a given subgraph, indicating how tightly interconnected a research community is. A dense citation graph suggests a mature, consensus-driven field where claims are heavily cross-validated. A sparse graph indicates a fragmented or nascent domain. Generative engines use density as a confidence calibration signal: claims originating from dense, highly cross-referenced clusters are assigned higher trust scores, while claims from isolated nodes or sparse regions face stricter factual grounding requirements before being surfaced.