Plagiarism Prevention Technology

The purpose of Plagiarism.org and Turnitin.com is to help put a stop to digital plagiarism. This means that papers will never again be recirculated or recycled every year, that papers will not be copied from one class and used for a different class, that papers from one university will not find their way to another university course, and finally, that papers acquired from the internet will never be used to fulfill a course requirement. We hope to achieve these results by offering educators the most advanced tools for plagiarism detection available anywhere. Now in its eighth year of development, our technology -- what we call Document Source Analysis -- uses a set of powerful algorithms to create a digital fingerprint of any text document. Here's how it works:

1. First, we make a digital fingerprint of any submitted document using a specially developed set of algorithms.
2. The document's fingerprint is cross-referenced against our local database containing hundreds of thousands of papers.
3. At the same time, we release automated web crawlers to scour the rest of the internet for possible matches.
4. Finally, we create a custom, color-coded originality report, complete with source links, for each paper.

Detection

When a manuscript is copied verbatim from a source on the Internet, it is easy for Turnitin.com to locate that source and flag the manuscript as unoriginal. However, some students may attempt to mislead an instructor by changing words, adding sentences, or reorganizing their papers. We've tested each of these methods to ensure that they do not thwart our anti-plagiarism technologies.

Numerous studies indicate student plagiarism is frequently the result of last-minute desperation, often precipitated by procrastination or simple laziness. In its most flagrant form, plagiarism occurs when a student copies a paper from another source without alterations of any kind. In such instances, Turnitin.com will immediately flag the paper and locate the original document. Less obvious examples occur when a student alters an existing work or integrates it, uncited, into his or her own. Some common examples include the substitution of words or sentences within an already existing work or the cutting and pasting of phrases or entire paragraphs from an outside source into a new document. Document Source Analysis has been optimized to ensure that papers plagiarized in this manner do not escape detection.

We have found, ironically, that the work involved in such an effort would be equal or greater to that necessary for the creation of a completely original document. Students attempting to bypass our system would quickly find they would be better off writing the paper themselves.

The sections that follow outline in detail how Document Source Analysis works to uncover even the most creative efforts of would-be plagiarists.

Word Substitution

Below is an example of a sophisticated attempt to replace specific words, in a paper on Macbeth, with a corresponding synonym. The upper excerpt is a portion of the original text; the lower paragraph is the same excerpt with the substituted words highlighted in red.

Macbeth Manuscript from the Internet (Intro Paragraph)

Macbeth is presented as a mature man of definitely established character, successful in certain fields of activity and enjoying an enviable reputation. We must not conclude, there, that all his volitions and actions are predictable; Macbeth's character, like any other man's at a given moment, is what is being made out of potentialities plus environment, and no one, not even Macbeth himself, can know all his inordinate self-love whose actions are discovered to be-and no doubt have been for a long time-determined mainly by an inordinate desire for some temporal or mutable good.

Same Manuscript with Modified Words

Macbeth is shown as an empowered man of well-established character, prosperous in several fields of life and enjoying an esteemed reputation. We mustn't conclude, therefore , that all of his volitions and actions will be foreseeable ; Macbeth's essence , like most other men at any given time, is what's being created out of potentialities and his environment, and no one, not even Macbeth himself, can discern all his immoderate self-love whose behaviors are found to be-and without doubt have been for some time-determined primarily by an extreme desire for a temporal or changeable good.

As you can see, students may attempt to find sensible synonyms for nearly half the words in a paper, but Turnitin.com will still flag the altered paper as such. Figure 1 (below) illustrates what happens as the percentage of word substitution in a paper increases. The red curve indicates our plagiarism index, or 'originality quotient'. It is scaled from 0.0 (no similarities)* to 1.0 (carbon copy). The yellow line indicates the threshold at which a paper has been altered beyond recognition. In this case, a student could replace about 45% of all words and the paper would still be considered plagiarized.

Figure 1. Plagiarism index vs. percentage of words altered.

When a submitted manuscript is contained within the Plagiarism.org database, there is no question it will be flagged in instances of plagiarism. Figure 2 (below) illustrates how an unoriginal manuscript sets itself apart from other papers in our database.

Figure 2. An exact duplicate of the Macbeth manuscript is flagged as unoriginal within one portion of our database.

After nearly half the words in the Macbeth manuscript were changed to synonyms, the paper was still flagged as being suspicious (Figure 3). In actuality this is an extreme and rather unrealistic example, for changing over half the words in a long essay would render any paper somewhat nonsensical, in addition to requiring an excessive amount of work.

Figure 3. The same Macbeth manuscript is still flagged after alteration.

Sentence Addition

The replacement or addition of new content is the most typical form of manuscript alteration. At what point does a manuscript with added or altered sentences or paragraphs become original?

The first paragraph (below) is an excerpt from an internet paper (A) on Macbeth. Below that is the same paragraph with new material cut-and-pasted directly into the body of the paragraph; the new material was from a second paper and is indicated in red.

Paper A

Macbeth Internet Derived Paper (Intro Paragraph)

Macbeth is presented as a mature man of definitely established character, successful in certain fields of activity and enjoying an enviable reputation. We must not conclude, there, that all his volitions and actions are predictable; Macbeth's character, like any other man's at a given moment, is what is being made out of potentialities plus environment, and no one, not even Macbeth himself, can know all his inordinate self-love whose actions are discovered to be-and no doubt have been for a long time-determined mainly by an inordinate desire for some temporal or mutable good.

Paper A + B

Macbeth Modified Test Paper with Combined Added Content

Shakespeare's famous play, Macbeth, is one of his great tragedies based around the classic theme of the hero's fatal flaw. Macbeth is presented as a mature man of definitely established character, successful in certain fields of activity and enjoying an enviable reputation. Yet, like any man, he is human, and thus in possession of flaw and foibles, hidden that they may be from public eye, and hinted at by foreshadow only by the author. We must not conclude, there, that all his volitions and actions are predictable; Macbeth's character, like any other man's at a given moment, is what is being made out of potentialities plus environment, and no one, not even Macbeth himself, can know all his inordinate self-love whose actions are discovered to be-and no doubt have been for a long time- determined mainly by an inordinate desire for some temporal or mutable good. This desire being so strong under certain circumstances as to override all others, even, as is usually the case in tragedy, the ultimate desire of self-preservation.

Figure 4 illustrates how a paper created in this manner is still easily detectable with Document Source Analysis. Although the plagiarism index drops gradually as higher percentages of the second paper are added to the first, it still remains an obvious case of plagiarism, even when the original document has been appended with up to 100% of the second text.

Figure 4. Plagiarism detection vs. percent of material added.

Figure 5 Shows how we flagged paper A as completely unoriginal when no new content was added.

Figure 5. Manuscript A is flagged as unoriginal within one portion of our database.

Even more noteworthy is the fact that combining two separate sources not only fails to escape detection, but actually increases the likelihood a paper will be flagged: instead of one source to detect, we now have two. Figure 6 illustrates how a work plagiarized in this manner will be caught by Document Source Analysis on both counts.

Figure 6. Our algorithms indicate that papers A and B were copied and used as the sources for the 'new' paper.

Tech Summary

Although it might seem improbable to expect high performance out of a search technology given the vast sea of written material out there, both in digital and conventional form, we feel that possible sources for a suspect paper can be narrowed down with a relatively high degree of accuracy. A thorough survey of the literature on term paper plagiarism has indicated that most instances of academic dishonesty involve the outright duplication of another work: from a colleague, from the internet, or even from oneself, by recycling an older paper from another class. Rarely will a student copy material directly from a book, out of fear that it will be known to the professor; studies have shown that even when this does occur it is, in fact, relatively easy for instructors to detect.

Even though we primarily compare papers from term paper mills and other university and high school classes, we are still forced to limit the search of our multi-gigabyte database* to those papers statistically related to the one in question. Through the use of proprietary methods of dimensional reduction, we are able to automatically group papers by categories and specific mathematical relationships. Figure 7 illustrates how a fraction of our database (blue dots) is separated from manuscripts that relate (red dots) to a target paper (black dot). The three-dimensional curve represents the probability that one of those red dots is a copy of the target paper; this system gives us the freedom to concentrate primarily on the 'red dot' papers.

Figure 7. Narrowing the Search

*Our database is constantly being updated by a series of automated web robots that scour the internet for rogue term papers. We also receive a continuous supply from the institutions and individual instructors using our service.