ai and plagiarism checker

ai and plagiarism checker

The Importance of AI in Plagiarism Detection

1. Introduction

In this paper, we explore the modern problem of plagiarism and the use of machine learning as a means of addressing this problem. In the near future, it is likely that an AI detection system will be implemented to further monitor students’ assignments. Therefore, it is important to be aware of the capabilities of such a system and the ethical implications of its use.

With the advent of modern computer technology, this practice can be monitored more effectively. By utilizing search engines and large databases, it is possible to identify instances of copied text by comparing a suspicious document against pre-existing documents. However, this method has its shortcomings. A simple change of words can sometimes fool the search engine, and a student may have a friend write a piece for them so that it is not available on the internet. Both of these are cases of plagiarism but are unlikely to be detected by this method. It would be desirable to have an automated system that can determine whether a given document is an original piece of work. This is the goal of machine learning, a field of artificial intelligence.

In recent years, with the development of the internet, it has become increasingly easy for students to “forget” to cite their sources, accidentally or intentionally, even if they know that it is wrong. Simultaneously, it has also become easier for teachers to catch the crime due to the existence of search engines, which are capable of retrieving vast amounts of information from various sources. Despite this, it is still a common practice among students and even professionals in the field, such as journalists.

2. Advantages of Using AI in Plagiarism Checking

The second advantage is the wide range of text that can be checked. There have been many cases in which human experts are unable to cross-reference a suspected text with the original source because it is too obscure. Usually, the experts use internet search engines, and more often than not, the search engine will not show the exact same text. Now, what if the text is in a foreign language? Manual detection is virtually impossible. If compared to AI, with enough resources, the AI can be tailored to cross-check specific text or a wide range of text from various sources and languages.

The first advantage of AI usage for plagiarism checking is the resources needed. Manual detection usually requires human experts whose availability cannot be guaranteed. Furthermore, the task is a bit mind-boggling. The human experts will usually print the suspected text and check it word per word on the internet search engine. The process usually takes a lot of time, and humans are prone to making mistakes. Depending on the amount of text, it is better to just pay a computer expert to program an AI specifically for plagiarism checking. Once the AI is finished, the checking is just a push of a button and the search suited to the AI’s skill.

Aristotle said, “When the internet came about, it was natural that people would take advantage.” Because the availability of information has become excessively easy and the interconnection is rapid, answering questions or writing articles can be promptly done using available text. It is likely that the work will become identical, and the original source will remain unknown. This is what we call plagiarism, an act of stealing other people’s thoughts and ideas manifested in the form of articles, essays, journals, reports, etc., without giving credit to the author. Now that AI progression is moving rapidly, there are many advantages of AI usage for plagiarism checking compared to manual detection.

3. Limitations of AI in Plagiarism Detection

Perhaps the largest limitation is the fact that a system’s detection of plagiarism will always be pitted against the human intellect. If a student can plagiarize material by skillfully evading detection through the use of paraphrasing or purposeful translation, it may always be the case that another human will need to identify this case of plagiarism in comparing both the student’s work and the original source. This presents an impasse for the automation of plagiarism detection and a question as to whether detecting plagiarism in all its forms is desirable.

Given the nature of AI systems, the system will only ever be able to check for plagiarized material against what has already been indexed within its databases. This could pose a problem for checking against student works as certain plagiarized materials can be repeatedly passed down and not necessarily originate from the original author. An intelligent plagiarizer may also try to trick AI systems by ‘watering down’ the extent of plagiarism to fall just below the detection threshold of the program.

A quoted material will be falsely indicated as plagiarized if quotation marks are not placed around the material, insufficient citations will also cause borrowed material to be flagged as plagiarism. Paraphrased content is also very difficult to detect as it is intentionally changed by the plagiarizer to evade detection. This presents a problem for current detection algorithms because there is no definitive method to check if material has been paraphrased, and changing a few words in a sentence or the sentence structure itself is considered sufficient as transforming the material into an original work. In the case of translated works, current detection tools are only able to identify content which has been directly translated.

4. Enhancing Plagiarism Checker with AI Technology

The use of semantic analysis is an effective way to prevent the act of plagiarism. Generally, plagiarism involves copying someone’s idea and then rephrasing it using different words. The original sentence may have an implicit meaning to the idea, and the person intending to commit plagiarism may not understand the context of the idea and rephrase the sentence to its literal meaning. A semantic analysis system will alert the user if it finds a similar sentence with ideas that have a close meaning. Enhancements can be made to the system to provide the user with alternative sentences that convey the same idea. This approach is proven to be very effective as it not only detects plagiarized sentences but also teaches the user the correct way to interpret the idea conveyed by the sentence.

Advancement in AI technology has the potential to develop a preventive tool to stop the act of plagiarism. The development of new processes of machine learning and natural language processing makes plagiarism prevention more viable. Several current research studies show that the use of AI technology can enhance the competency of machines in paraphrasing articles. This process involves translating the original sentence into a new sentence with the same meaning. However, this is proven to be a difficult task for machines due to the complexity of the semantic meaning in a sentence and the use of idiomatic expressions. Machines are trained with algorithms to identify the contextual meaning of the original sentence.

5. Conclusion

The industry of plagiarism detection in the field of programming and computer science is far from saturated. Automated tools have provided a first great step, however in this area in particular the discrepancy is still quite large between automated and human understanding of copied code. More intelligent techniques targeted at the identification of methods and algorithms rather than simple text similarity are required if we are to empower educators with a viable alternative to manual detection, and still provide students with a learning experience from mistakes. Finally, formalised methods and tools for comparison and identification of copied code can provide an alternative cost effective solution for copyright and intellectual property dispute in the software industry. As authorship of software is a growing market, there is much potential for direct comparison tools to be used for the purpose of litigation between two bodies of code.

We have developed techniques for plagiarism detection, which are now largely effective. However, by placing more emphasis on hardcoded text, the issue of decompiled code plagiarism still looms over the software engineering and computer science course at UNSW. We believe that with further work in this area, we can solve this problem and result in a reduction of copied code by students. With a move to electronic submission methods increasing, the tools and techniques outlined in this report can provide benefit to manual methods currently used for comparison of student code.