Searching for Similar Files

The content of files in the IS UCP data repository, for which a plain text version is available, is continuously machine-analyzed. Each document (final work, seminar work, essay, presentation, ...) that the user finds in the system can be compared and displayed with resources that have a similar text:

• click on the line with the file (right-click will bring up the details panel on the right, left-click will bring up the context menu),
• use the "Find similar documents" operation (two eggs icon),
• use the displayed similarities (or adjust the settings and have the similarities recalculated) to assess whether it may be plagiarism or not.

For the documents found to be similar, IS UCP displays the similarity level in percentages. Each document also provides a graphical interpretation of similarities in the form of a linear or 2D map:

The user can smoothly switch between detected similarities and detailed comparison of two documents using the magnifying glass button on the selected document. It is possible to immediately view any similar document found without losing context of the original document. Comparison view of two documents:

The system searches for similarities across the shared database of compared documents, which includes, in addition to the IS UCP document server, final theses of participating schools in the Theses.cz system, seminar and other theses in the Odevzdej.cz, system, scientific publications in the Repozitar.cz system and other documents in information systems operated by MU.

Because a significant amount of work and documents are not public, searching the common database is an effective tool for finding possible plagiarism. A student cannot just submit a seminar paper to a friend from another school, because there is a high probability that he is already in the system and there would be similarities. The assignment of seminar and final thesis topics is often repeated, and it is therefore likely that the system will already know about the source documents from which this topic is usually drawn.

The search for similarities also includes an algorithm that analyzes the document being compared and examines it from sources on the Internet. This procedure has various technical limitations and is computationally and time consuming. Therefore, it is preferably used for archives of final theses. Due to the speed of processing, seminar papers in submission offices and long answers in ROPOTs are primarily compared against a shared database of all source documents. This includes, among other things, resources from the Internet, which have been traced in the past for the purpose of checking some of the millions of final theses being compared.

However, there is a difference between finding similarities to seminar papers (for which the shared resource database is used) and final papers (for which the shared resource database is enriched with newly downloaded resources from the Internet). But there is no need to worry, the database is really large, so in most cases the system "knows" enough resources that students usually draw from, as well as enough resources that students usually describe.

The supervisor, who usually wants to check the quality of the work when writing the report, proceeds with the help of

and select "Find similar documents". They will see the overall similarity percentage and highlighted matches, which they can go through step by step using the arrow buttons to assess whether the passages are properly cited.

If you have a document in the form from which a text version can be created (it is not a picture, for example), insert it anywhere in the IS UCP (eg in the Provider, in the Study Materials of the subject or in My website) and IS UCP in the similarity search mechanism includes.

Then use the “Find similar documents” operation (the icon with two eggs). For more details, see the question How are text similarities (suspected plagiarism) revealed in IS UCP?

By default, documents uploaded by the same uploader are not compared. Therefore, you need to click the gear icon in the top toolbar to open Settings, and then select the option “include documents of the same uploader” and display them:

This is a comparison of documents with each other:

• A text searchable version is generated for each document in the database. In this text version, the algorithm focuses on similar or paraphrased sections of text and evaluates the degree of agreement across the entire shared document database, including Internet resources.
• Texts in Czech, English and Slovak are compared, the condition is that they have at least a few sentences or paragraphs (in very small files there is not enough text for their analysis and finding similarities).
• Before the result is presented to the user, those documents that overlap only in passages that are the same as for previously found sources are omitted. Practically, for example, it is a citation of a certain law in another hundred final theses and documents on the Internet. If there are less than 10 similar sources, they will all be displayed without omission.
• The user is shown the most relevant documents that have a significant similarity to the document being searched, and the percentage of that similarity.

If students copy from each other, the system evaluates their answers as similar and displays the percentage of similarity. You can find more on this topic in the question How works compare similarities?

Authors can verify that their text has not been used in another author's work.

What the similarity search interface shows and how to use it:

For the documents found to be similar, IS UCP displays the similarity level in percentages. Each document also provides a graphical interpretation of similarities in the form of a linear or 2D map:

The user can smoothly switch between detected similarities and detailed comparison of two documents using the magnifying glass button on the selected document. It is possible to immediately view any similar document found without losing context of the original document. Comparison view of two documents:

Some similar documents do not provide information about the new match, only recurring similar passages of the displayed source documents. The system skips these documents. In practice, for example, these are more extensive citations of a certain law in another hundred final theses and documents on the Internet. Therefore, users will now not see all documents in which copyright law paragraphs are quoted, but will see only one of them. The one that has the most other similarities with the selected document.

First of all, you need to see if you just see another document that contains the similarity. For example, whether the system did not mark an equal collection of laws with the text of the law, and not a newspaper article commenting on the same article.

The system also checks longer texts entered as answers to ROPOTs. Answers to questions of the type "Enter text" (: a) are checked if more than 50 characters were entered. The system automatically inserts these answers into the Study Materials of the subject if the ROPOT is in the study materials. If the ROPOT is not in the study materials, then it is stored under the ROPOT description file. These files are not available to students by default.

The following can be used to check the answers in the ROPOTs:

The individual files, student names and similarity percentage are displayed. You can click on the similarities for each file (ie the individual long answers in the ROPOT) for a more detailed examination.

The similarity search tool can help prevent copying or resubmission of senior classmates' essays. If students copy from each other, the system evaluates their answers as similar and displays the percentage of similarity. To check the work in the delivery room, use:

The individual files, student names and similarity percentage are displayed.

The “Similar Documents” application now naturally includes a function for comparing two selected documents. In the list of similar documents, a selected document can be displayed together with the examined document by clicking the button with the magnifying glass icon. The correlation between similar texts in both documents is displayed visually at positions where similarities were found, including the degree of similarity of the identified passages.

The application functions even if the user does not have access to the source document. In such cases, only the beginnings of similar words from the source document are shown, and the rest of the text is intentionally obscured. This serves as guidance for at least a rough assessment of the seriousness of the textual similarity.

The application includes a number of graphical elements that make it easier to navigate the identified similarities. The degree of similarity of a passage is indicated by the color and fill level of a circle in various parts of the interface:

Both the examined and source document are displayed in a similar way. The text is divided into shorter parts, approximately paragraph length, and two consecutive paragraphs from the examined document are compared with three paragraphs from the source document.

If the user wants to compare two particular documents directly and is concerned that one of them might otherwise be skipped due to low significance, they can use the application for direct comparison of two specifically selected documents. Both documents must be uploaded in the IS UCP.

The 2D map visually displays the location of similar passages in both documents.

For example, similarities in the top-right corner often represent matches between the concluding sections of the documents (typically bibliographies).

The 2D map provides a quick overview of the nature of the similarities. Some examples:

Strong diagonal

A strong red diagonal indicates a long continuous passage found in both documents. The example shows a dissertation made up of several published articles, one of which is the source document. Other similarities indicate longer domain-standard expressions used in other articles forming the dissertation. Note: the diagonal angle may differ depending on the relative lengths of the documents.

Author’s introductory statement

A short similarity in the bottom-left corner typically represents an author’s acknowledgement or declaration, often standardized within an institution.

List of references

A cluster in the top-right area represents longer text sequences in the bibliography. This usually indicates that the works rely on similar sources and does not constitute plagiarism. A roughly diagonal direction suggests that references are listed in a similar order, sorted by the same criterion (e.g., year of publication).

Several dots aligned vertically

This means a single passage in the compared document corresponds to multiple places in the source document (or vice-versa horizontally). Typically these are common domain phrases, definitions, or standard terminology. Usually not a concern.

Low-intensity similarity

If there are no strong clusters or diagonal lines, and similarities are mostly yellow or light orange, it likely reflects common phrasing or shared subject matter rather than plagiarism.

Rephrased text

This is a true case of plagiarism: an entire work was rewritten with significant rephrasing of another text. The map contains few red dots, but the wavy diagonal pattern indicates near-continuous similarity. Gaps in the diagonal may mean undetected similarity or sections sourced from another document. Dots along the top edge represent references, where the source cites literature only at the end, while the compared document cites throughout (e.g., via footnotes).

Copied chapter

Same case as above, but here the source is a Wikipedia article. The plagiarized text follows the same structure but rephrases and omits parts. Due to the large length difference between documents, the diagonal slope is almost vertical. Even approximately two pages of copied text are significant and worth addressing.

Introductory section

In many fields, introductory sections summarize background knowledge. Here, two secondary-school theses of similar length from the same school clearly drew on the same sources (or each other) in the introduction. The second halves (right/top of map) show no similarity, suggesting the original work begins there.

Plagiarism (copying, imitations) is an unethical activity and fraud that has no place in college. Acceptance of the text of another author without a visible indication of the copied text with the exact citation of the source constitutes an unauthorized infringement of copyright in violation of copyright law. There is a risk of sanctions for the offense, even after graduation.