Diagnosing pelagism using graphs in Persian texts

Master's Thesis in Computer Engineering

Trend: Software

Abstract

This thesis focuses on graph-based similarity search in natural language texts. The need for a strong method to present texts is an important issue in the field of plagarism detection. In this project, according to this need, we have introduced a powerful method to present natural language and have used it to detect plagarism. For this purpose, we have defined the concept of "graph modification distance" and used it to calculate the distance between two graphs. Sentences are represented by dependency graphs in which words are connected by their dependencies. Dependency graph extracts the grammatical structure of sentences. The graph-based similarity method has been used in the problem of pelargism detection. The main advantage of graph-based presentation is related to the ability of this method to detect similarities between words. The evaluations showed that the results obtained from the dependence graph have better results than the methods of direct comparison of the graphs. Using the graph modification distance to compare two graphs leads to the improvement of the dependency graph results and increases its efficiency. Keywords: graph modification distance, natural language processing, dependency graphs, detection of plagarism. Chapter 1. Introduction. Some people may intentionally or unintentionally use the works of researchers while not even mentioning the name of the main author of the work, this act is called plagarism. can be Plagarism is the intentional or unintentional act of copying or using the ideas and works of others without crediting the original source. Based on the investigations conducted on the issue of plagarism and the number of articles published on the web and the reflection of concern about its occurrence, it is clear that most educational and research centers in their electronic pages in all parts of the world, including in the developing countries of Asia and Africa, under the influence of the publishers with a history of research journals and also to preserve the scientific dignity of their centers, have started to comprehensively introduce this anti-ethical and anti-social phenomenon as a crime. Plagarism is a problem in the scientific community and is growing rapidly, because data and information from electronic documents and the Internet are quickly and easily obtained through copying and pasting from those sources. This problem occurs when the content of illegal documents is found without permission and without citation. This problem is known as plagarism, and plagarism can include a wide range of deliberate manipulations to accidental copying of other people's content. The main goal of this thesis is to express the graph-based method for text presentation and its use in pelarism detection. The following sections of this chapter state the reasons for using this method and describe the problem of pelargism diagnosis. In addition, they provide a solution for this problem. Finally, the structure of the introductory thesis and the topics that will be discussed in the next chapters will be reviewed.

The problem of finding similarities between two texts is a common problem in the field of natural language processing. In order to evaluate the similarities between two texts, each text needs a method of presentation. One method is simple text, where a list of words forms a sentence. Plain text is usually used because of its simplicity, but it lacks clear information about the grammatical structure.

Some aspects of language are better presented using structural presentation methods such as dependency graphs that include words connected together.

Dependency graphs extract the grammatical structure of a sentence, and are limited to the scope of the same sentence. One of the main advantages of graph-based representation is that dependency graphs are in most cases insensitive to word order. This advantage makes it possible to find similarities between sentences in which the word order is scrambled.

A stronger representation provides a better foundation for identifying similarities in complex situations. Consider the two sentences presented in the dependence graph of Figures 1-1 and 1-2. Humans must be able to determine that these sentences have the same meaning. However, the automatic recognition of the similarity of these two sentences will have problems due to the substitution of words. If plain text presentation is used, the only common words will be "to", "out", "from", "shot" and "ball". The words "shot" and "ball" are the only words that convey the meaning of the sentence.By looking closely at the dependency graphs, it is clear that there is structural similarity between the sentences.

With the increasing availability of texts on the web, plagarism has become simpler and easier. A large amount of plagarism texts in the field of teaching and learning are increasing year by year. As a result, there is a strong need for automatic detection of plagarism. 1-1 Problem description The most important part of this thesis is the implementation of an algorithm to calculate the graph modification distance, which calculates the similarity between two graphs. The algorithm is based on the calculation of the number of editing operations required to transform one graph into another [1]. Each editing action has an editing cost, which determines how much an action costs.

Automatic plagarism detection is a research field that is basically based on textual similarity. The problem of using graph-based text similarity to detect plagarism is stated in research question 1. Research question 1: Is graph-based similarity, within a certain graph modification distance, applicable and computationally feasible for plagarism detection? Experimental plagarism detection systems are often based on simpler text representations, such as ngram matching and vector space model [2 and 3]. As a result, the method can be relatively unique. Due to the uniqueness of the method, some implementation details are not defined. Research question 2 states a problem that specifies the details of the algorithm. Research question 2: What is the best method for calculating the graph editing distance between sentences, especially in terms of the cost of editing and presenting the graph in the field of plagarism detection? In order to evaluate the efficiency of the algorithm for calculating the graph editing distance, its efficiency is compared to existing test systems. Research question 3 raises the problem of comparing the algorithm with existing methods. Research question 3: How does graph-based similarity compare with other methods in the detection of plagarism, such as index-based recovery and ngram matching? 1-2 solutions The following objectives guarantee the answers to the questions raised in the previous section. It will be implemented on the test graph editing interval. The system should be able to evaluate many texts and recognize plagiarism at the sentence level.

Research objective 2: Some functions will be presented to calculate the cost of implementation editing. Different methods are compared with each other and the resulting systems are evaluated.

Research objective 3: The output of the experimental system is compared with existing pelargism detection systems. This comparison is considered an experimental evaluation of the system. 1-3 problems in the implementation of the algorithm One of the most important problems in the implementation of semantic plagarism detection software in the Persian language is related to the Persian language itself. As we know, the Persian language is a very complex language and has a grammar with many exceptions, and due to its similarities with the Arabic language, it has problems with the characters that are common in Arabic and Persian, which causes some cases to be misdiagnosed. The Persian language, which is almost similar to the Arabic language, has various forms and changes depending on the third person. These exceptions and thousands of similar problems due to the rich and complex grammar of the Persian language cause the task of detecting plagarism to face many problems and the need to consider many aspects in the implementation of the algorithm.

Another problem that may arise in the field of semantic plagarism detection is that it is not enough to only use the dictionary of synonyms of words, because it is possible that users who engage in the inhuman act of plagarism and copying other people's works, in some cases instead of replacing the synonyms instead of their originals, with a slight change in the structure and appearance of the sentences, use the opposites of the words instead of their originals. Sometimes even without replacing the words, it is possible to create similar sentences with a slight change in the structure of the sentences.