How Originality Generates Similarity Reports

In this article:

• Report generation pipeline
• Semantic similarity
• Indexing and the do-not-index signal
• Worked examples
• Supported file size and formats
• Originality boundaries

Report generation pipeline

A document moves through Originality in the following stages:

1. Receipt. A document arrives at Originality from an integrating system (WISEflow, Canvas, Moodle, Digital Eksamen, or a custom system using the API).
2. Indexing decision. Originality checks whether the document should be added to the comparison database. This is controlled by the do-not-index signal sent by the integrating system (read more below).
3. Analysis. The document is compared against the comparison database (from your own licence and sharing groups, if applicable) and against indexed web sources.
4. Report creation. A similarity report is generated within 24 hours of receipt and is returned to the integrating system.
5. 24-hour update window. If further documents arrive within the next 24 hours, the report is updated to include any new potential matches. After this window, the similarity report is locked in.

Cross-document matches: Two documents sent to Originality within 24 hours of each other can match against one another. In that case, each report will reference the other document as a match.

Document transfer to Originality: Some integrating systems deliberately hold submissions for a period before sending them to Originality, to accommodate late submissions or withdrawals. For example, WISEflow holds submissions for two hours after the participation period closes. Read the relevant integration article to learn more about the timing of your platform.

Semantic similarity

Originality does not rely on exact string matching. It uses a language model and a vector database to compare the meaning of sentences rather than just the surface form. This means Originality can identify:

• Exact text matches
• Minor edits and word substitutions
• Paraphrased sentences that retain the original meaning

Matches are categorised by match strength, which reflects how closely the two sentences correspond.

Indexing and the do-not-index signal

Each document received by Originality is either added to the comparison database (indexed) or not added. Indexed documents become a comparison source for future submissions. Non-indexed documents are analysed and produce a report, but never appear as a match for anyone else.

Originality respects a do-not-index signal on each incoming document. This signal is typically driven by a declaration the participant makes when submitting their paper – for example, that the work contains confidential material. How the signal is collected varies by the integrating system:

• WISEflow: The participant ticks Confidential Material on the cover sheet.
  

• Canvas/Moodle/Digital Eksamen: Collected via the connector's submission flow, where supported.
• Custom systems via the API: The do-not-index signal is passed as a parameter on the upload request.

Worked examples

Example 1 – Exact match

• Source: "The mitochondrion is the powerhouse of the cell."
• Submission: "The mitochondrion is the powerhouse of the cell."
• Result: Exact match.

Example 2 – Paraphrase

• Source: "The mitochondrion is the powerhouse of the cell."
• Submission: "Cells rely on the mitochondrion as their main source of energy."
• Result: Moderate match, typically lower strength than an exact match but still flagged.

Example 3 – Common phrase, excluded

• Submission: "In conclusion, the results show..."
• Result: Excluded automatically as a short common phrase (see the Excluded tab in the report).

Supported file size and formats

Originality boundaries

• Originality does not make a plagiarism determination – this is a task for the assessor. What Originality does is it supports the assessor in the reviewing process and identifies document similarity.
• Originality does not produce a single "originality percentage" that is meaningful in isolation – the document score must be read alongside the matches themselves.
• Originality does not compare against material outside its indexed sources (indexed sources are your own licence documents, sharing group documents, and indexed web content).