Summary controls

Iteration playback

Graph & ranking stats

Sentences (N)—
Graph edges—
Iteration0
Converged at—
Last Δ (max)—
Score range—
Summary sentences—

Top-ranked sentence

Index—
TextRank score—
—
TextRank builds a graph where every sentence is a node and edge weights are the overlap-based similarity between sentences. Scores update by power iteration of a damped PageRank recurrence: score(i) = (1−d)/N + d·Σⱼ w(j,i)/Σₖw(j,k)·score(j). The top-scoring sentences, re-ordered to match the original document, form the extracted summary.