How to Reduce Similarity Index/Plagiarism Below 10% || Hindi

A similarity index above the usual thesis threshold can be driven less by “plagiarism” and more by avoidable overlap—especially when citations, quotation marks, and filter settings aren’t handled correctly. The practical target laid out here is bringing overall similarity to “below 10%” (and keeping it within an “up to 10%” range) so the work passes evaluation expectations, even when earlier checks show figures like 45%.

The first lever is how the similarity report is interpreted and configured. Similarity percentages are not a direct measure of plagiarism; evaluators typically look at how much of the flagged text overlaps with external sources, then judge whether that overlap is properly attributed. The transcript emphasizes that the report’s total is built from multiple categories—such as “not cited” and “not quoted” segments—combined into one number. That means a single oversight (for example, leaving a quoted or cited passage without quotation marks, or failing to cite a source even when the idea is borrowed) can inflate the overall score. It also stresses that the software’s filter settings must match thesis rules; otherwise, the report can overcount items that should be excluded.

Next comes a set of exclusion filters tied to common thesis guidelines. For thesis-style work, quoted material can often be excluded from similarity calculations, but only when permissions exist and the material is properly cited. The transcript also lists other typical exclusions: bibliography/references, table of contents, preface, acknowledgements, and generic elements like standard symbols, standard equations, and standard terminology. It warns that excluding “cited text” blindly can be risky because the goal is to keep meaning intact while ensuring the original source is still credited.

The transcript then focuses on the mechanics of reducing similarity through writing practices. Proper paraphrasing is treated as the main “weapon”: instead of copying and swapping a few words with synonyms, the writer should restate the idea in their own wording while preserving the original meaning and maintaining correct citation. It highlights a common failure mode—changing synonyms without changing word order or grammar—because many plagiarism/similarity tools detect matches based on sequences and structure. A safer approach described is to read the source, then reorganize and rewrite the full sentence or paragraph in one’s own flow, optionally using active/passive voice changes, while keeping non-changeable elements (like abbreviations or common knowledge) consistent.

Finally, the transcript recommends using tools and workflow habits to prevent reference-related misses: use plagiarism-checking and “authenticity”/detection features to identify what was borrowed, maintain detailed notes that track which source each idea came from, and use reference management software (e.g., Mendley, Zotero) to avoid missing citations. The closing strategy is to target the external-source matches individually—aiming to reduce each flagged source’s contribution (e.g., from around 4% down below 1%)—because that naturally pulls the overall similarity below 10%.