Plagiarism Word Percentage Calculator (Two Texts)
Calculate the percentage of matching words between two texts to get a quick similarity estimate.
Reviewed by Daniel Agrici, Founder & Lead Developer
Formula
Similarity % = (Matching Words / Total Words) x 100; Jaccard = |A intersect B| / |A union B| x 100
Word overlap counts matching word instances across both texts. Jaccard similarity uses unique word sets, dividing the intersection (words in both) by the union (words in either). Bigram similarity compares consecutive word pairs for phrase-level matching.
Worked Examples
Example 1: Detecting Direct Copying
Problem:Text A: 'The quick brown fox jumps over the lazy dog near the river.' Text B: 'The quick brown fox leaps over the lazy dog by the river bank.' Compare similarity.
Solution:Text A words (min 3 chars): the, quick, brown, fox, jumps, over, the, lazy, dog, near, the, river = 12 words\nText B words: the, quick, brown, fox, leaps, over, the, lazy, dog, the, river, bank = 12 words\nMatching words: the(3), quick(1), brown(1), fox(1), over(1), lazy(1), dog(1), river(1) = 10 matches\nSimilarity A: 10/12 = 83.3%\nSimilarity B: 10/12 = 83.3%
Result:Average Similarity: 83.3% | High Risk | 8 unique words in common
Example 2: Comparing Paraphrased Content
Problem:Text A: 'Machine learning algorithms analyze patterns in data to make predictions.' Text B: 'Artificial intelligence systems examine trends in datasets to forecast outcomes.' Compare similarity.
Solution:Text A words (min 3 chars): machine, learning, algorithms, analyze, patterns, data, make, predictions = 8 words\nText B words: artificial, intelligence, systems, examine, trends, datasets, forecast, outcomes = 8 words\nMatching words: 0 (completely different vocabulary)\nJaccard similarity: 0/16 = 0%
Result:Average Similarity: 0% | Low Risk | Successful paraphrasing detected
Frequently Asked Questions
How does word-based plagiarism detection work?
Word-based plagiarism detection works by comparing the vocabulary and word patterns between two texts to determine similarity. The process begins by tokenizing both texts into individual words, normalizing them to lowercase, and removing punctuation. Then the algorithm counts how many words appear in both texts, calculating the percentage of shared content. More sophisticated methods also analyze word frequency, n-gram patterns (consecutive word sequences), and positional information. This approach catches direct copying and close paraphrasing but may miss structural plagiarism where ideas are restated with entirely different vocabulary. Word-based detection is a first-pass method that works well for identifying obvious copying and provides a quantitative similarity score for further investigation.
What is a safe similarity percentage for academic papers?
Acceptable similarity percentages vary by institution and context, but general guidelines exist. Most universities consider below 15 to 20 percent similarity as acceptable for academic papers, accounting for common phrases, technical terminology, and properly cited quotations. Between 20 and 40 percent similarity raises concern and typically triggers closer review. Above 40 percent similarity is generally considered problematic and may indicate significant plagiarism. However, context matters enormously. A literature review section may legitimately have higher similarity due to quoted sources. Scientific papers using standard methodologies may share common phrases. Properly cited direct quotes increase the percentage without constituting plagiarism. Many institutions use Turnitin or similar tools and focus on the similarity report details rather than just the overall percentage.
What is the difference between Jaccard similarity and word overlap percentage?
Jaccard similarity and word overlap percentage measure text similarity differently. Word overlap percentage counts total matching word instances (including duplicates) divided by the total word count of one text. This means if a word appears 5 times in both texts, it counts as 5 matches. Jaccard similarity instead looks only at unique vocabulary, calculated as the number of unique words shared by both texts divided by the total number of unique words across both texts combined (the union). Jaccard ranges from 0 to 100 percent. Jaccard is less sensitive to text length differences and word repetition, making it better for comparing texts of very different lengths. Word overlap is more sensitive to actual content volume that matches and better detects verbatim copying of passages.
What are n-grams and why are they important for detecting plagiarism?
N-grams are sequences of N consecutive words extracted from a text. Bigrams (2-grams) are pairs of consecutive words, trigrams (3-grams) are triplets, and so on. N-gram analysis is important for plagiarism detection because it captures word order and phrasing patterns, not just vocabulary overlap. Two texts might share many individual words but have low n-gram similarity if the words appear in different contexts and orders. High bigram or trigram similarity strongly suggests that passages were copied because the probability of two independently written texts sharing the same multi-word sequences is very low. Professional plagiarism detection tools typically use 4-grams or 5-grams as their primary matching unit, as these are long enough to be distinctive while short enough to catch paraphrased copies.
References
Reviewed by Daniel Agrici, Founder & Lead Developer · Editorial policy