Academic Integrity & Text Attributions Explorer

Comparative Forensic N-Gram & Citation Alignment Workbench

Forensic Classification Analyzing... Attribution taxonomy match
Max Verbatim Run 0 Consecutive identical words
N-Gram Overlap 0.0% Matching sliding windows
Token Overlap & Jaccard 0.0% 0 shared unique lemmas

1. Canonical Reference Literature

0 tokens
Source Visual Text Inspection:

2. Evaluated Manuscript Draft

0 tokens
Manuscript Alignment Heatmap:

Academic Integrity Taxonomy & Evidence Audit

Standardized forensic diagnostic criteria for research ethics reviews

Pending Audit
Diagnostic Metric Calculated Level Integrity Threshold Compliance Assessment
Max Verbatim Sequence 0 words > 6 consecutive words require direct quotation marks Compliant
N-Gram Window Overlap 0.0% > 25.0% indicates substantial unoriginal sequence borrowing Compliant
Token Containment Index 0.0% Ratio of suspect vocabulary derived directly from source Low Borrowing
Attribution Formal Citation Absent Mandatory attribution tag or footnote required Citation Missing

Text overlap, denominators and checkbox labels

Read the explanation

Use two harmless example texts: red blue green gold, and red blue green silver. This tokenizer lowercases ASCII letter and digit words and excludes punctuation from word metrics. Three unique words are shared. The union contains five, so Jaccard similarity is three divided by five, or sixty percent. The comparison contains four unique words, so containment is three divided by four, or seventy five percent. The bars use eighty pixels per unique word. A two hundred forty pixel shared portion belongs to a four hundred pixel union or a three hundred twenty pixel comparison. Different denominators produce different percentages. Neither is a probability that misconduct occurred. The native slider starts at a minimum span of three words. Each four word text has two three word windows. Red blue green matches, while blue green silver is absent from the source. One matching comparison window out of two gives fifty percent overlap. The longest contiguous shared run is three words. The first half of the bar therefore fills on a shared two window scale. The source stores unique source windows in a set, but counts matching comparison windows by position. Repeated windows can therefore contribute repeatedly. If a text is shorter than the selected span, the denominator is floored to one rather than producing a division by zero. These are local string comparisons, not a semantic reading or an authorship investigation. Without the citation checkbox, the example receives the app moderate label because fifty percent overlap exceeds its thirty percent threshold. The longest run of three is below its separate five word threshold. The number line uses four pixels per percentage point, marking thirty, fifty and sixty five at their true positions. Checking citation switches to an attributed label while every numerical metric remains unchanged. The boolean branch does not inspect a real citation, quotation validity or permission. The native example test confirmed this exact label change. Treat the output as authored heuristic bookkeeping that requires human source review, not evidence for discipline or a factual accusation.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.