Krisnawati, Lucia Dwi
Plagiarism detection for Indonesian texts.
Dissertation, LMU München: Faculty for Languages and Literatures
As plagiarism becomes an increasing concern for Indonesian universities and research centers, the need of using automatic plagiarism checker is becoming more real. However, researches on Plagiarism Detection Systems (PDS) in Indonesian documents have not been well developed, since most of them deal with detecting duplicate or near-duplicate documents, have not addressed the problem of retrieving source documents, or show tendency to measure document similarity globally. Therefore, systems resulted from these researches are incapable of referring to exact locations of ``similar passage'' pairs. Besides, there has been no public and standard corpora available to evaluate PDS in Indonesian texts.
To address the weaknesses of former researches, this thesis develops a plagiarism detection system which executes various methods of plagiarism detection stages in a workflow system. In retrieval stage, a novel document feature coined as phraseword is introduced and executed along with word unigram and character n-grams to address the problem of retrieving source documents, whose contents are copied partially or obfuscated in a suspicious document. The detection stage, which exploits a two-step paragraph-based comparison, is aimed to address the problems of detecting and locating source-obfuscated passage pairs. The seeds for matching source-obfuscated passage pairs are based on locally-weighted significant terms to capture paraphrased and summarized passages. In addition to this system, an evaluation corpus was created through simulation by human writers, and by algorithmic random generation.
Using this corpus, the performance evaluation of the proposed methods was performed in three scenarios. On the first scenario which evaluated source retrieval performance, some methods using phraseword and token features were able to achieve the optimum recall rate 1. On the second scenario which evaluated detection performance, our system was compared to Alvi's algorithm and evaluated in 4 levels of measures: character, passage, document, and cases. The experiment results showed that methods resulted from using token as seeds have higher scores than Alvi's algorithm in all 4 levels of measures both in artificial and simulated plagiarism cases. In case detection, our systems outperform Alvi's algorithm in recognizing copied, shaked, and paraphrased passages. However, Alvi's recognition rate on summarized passage is insignificantly higher than our system. The same tendency of experiment results were demonstrated on the third experiment scenario, only the precision rates of Alvi's algorithm in character and paragraph levels are higher than our system. The higher Plagdet scores produced by some methods in our system than Alvi's scores show that this study has fulfilled its objective in implementing a competitive state-of-the-art algorithm for detecting plagiarism in Indonesian texts.
Being run at our test document corpus, Alvi's highest scores of recall, precision, Plagdet, and detection rate on no-plagiarism cases correspond to its scores when it was tested on PAN'14 corpus. Thus, this study has contributed in creating a standard evaluation corpus for assessing PDS for Indonesian documents. Besides, this study contributes in a source retrieval algorithm which introduces phrasewords as document features, and a paragraph-based text alignment algorithm which relies on two different strategies. One of them is to apply local-word weighting used in text summarization field to select seeds for both discriminating paragraph pair candidates and matching process. The proposed detection algorithm results in almost no multiple detection. This contributes to the strength of this algorithm.