Due to rapid growth of research articles in various languages, cross-lingual plagiarism detection problem has received increasing interest in recent years. Cross-lingual plagiarism detection is more challenging task than monolingual plagiarism detection. This paper addresses the problem of cross-lingual plagiarism detection (CLPD) by proposing a method that combines keyphrases extraction, monolingual detection methods and machine learning approach. The research methodology used in this study has facilitated to accomplish the objectives in terms of designing, developing, and implementing an efficient Arabic – English cross lingual plagiarism detection. This paper empirically evaluates five different monolingual plagiarism detection methods namely i)N-Grams Similarity, ii)Longest Common Subsequence, iii)Dice Coefficient, iv)Fingerprint based Jaccard Similarity  and v) Fingerprint based Containment Similarity. In addition, three machine learning approaches namely i) naïve Bayes, ii) Support Vector Machine, and iii) linear logistic regression classifiers are used for Arabic-English Cross-language plagiarism detection. Several experiments are conducted to evaluate the performance of the key phrases extraction methods. In addition, Several experiments to investigate the performance of machine learning techniques to find the best method for Arabic-English Cross-language plagiarism detection. According to the experiments of Arabic-English Cross-language plagiarism detection, the highest result was obtained using SVM   classifier with 92% f-measure. In addition, the highest results were obtained by all classifiers are achieved, when most of the monolingual plagiarism detection methods are used.

Identifying of compound nouns is important for a wide spectrum of applications in the field of natural language processing such as machine translation and information retrieval. Extraction of compound nouns requires deep or shallow syntactic preprocessing tools and large corpora. This paper investigates several methods for extracting Noun compounds from Malay text corpora. First, we present the empirical results of sixteen statistical association measures of Malay <N+N> compound nouns extraction. Second, we introduce the possibility of integrating multiple association measures. Third, this work also provides a standard dataset intended to provide a common platform for evaluating research on the identification compound Nouns in Malay language. The standard data set contains 7,235 unique N-N candidates, 2,970 of them are N-N compound nouns collocations. The extraction algorithms are evaluated against this reference data set. The experimental results  demonstrate that a group of association measures (T-test , Piatersky-Shapiro (PS) , C_value, FGM and  rank combination method) are the best association measure and outperforms the other association measures for <N+N> collocations in the Malay  corpus. Finally, we describe several classification methods for combining association measures scores of the basic measures, followed by their evaluation. Evaluation results show that classification algorithms significantly outperform individual association measures. Experimental results obtained are quite satisfactory in terms of the Precision, Recall and F-score.

No abstract

With the exponential growth of textual information available from the Internet, there has been an emergent need to find relevant, in-time and in-depth knowledge about crime topic. The huge size of such data makes the process of retrieving and analyzing and use of the valuable information in such texts manually a very difficult task. In this paper, we attempt to address a challenging task i.e. a crawling and classification of crime-specific knowledge on the Web. To do that, a model for online crime text crawling and classification is introduced. First, a crime-specific web crawler is designed to collect web pages of crime topic from the news websites. In this crawler, a binary Naive Bayes classifier is used for filtering crime web pages from others. Second, a multi-classes classification model is applied to categorize the crime pages into their appropriate crime types. In both steps, several feature selection methods are applied to select the most important features. Finally, the model has been evaluated on manually labeled corpus and also on online real world data. The experimental results on manually labeled corpus indicate that Naive Bayes with mutual information and odd ratio feature selection methods can accurately distinguish crime web pages from others with an F1 measure of 0.99. In addition, the experimental results also show that the Naive Bayes classification models can accurately classify crime documents to their appropriate crime types with Macro-F1 measure of 0.87. Our results also on online real word data show that the focused crawler with two-level classification is very effective for gathering high-quality collections of crime Web documents and also for classifying them.