Qasem A. Al-Radaideh

Using N-grams for Arabic text searching

N-grams have been widely investigated for a number of text processing and retrieval applications. This article examines the performance of the digram and trigram term conflation techniques in the context of Arabic free text retrieval. It reports the results of using the N-gram approach for a corpus of thousands of distinct textual words drawn from a number of sources representing various disciplines. The results indicate that the digram method offers a better performance than trigram with respect to conflation precision and conflation recall ratios. In either case, the N-gram approach does not appear to provide an efficient conflation approach due to the peculiarities imposed by the Arabic infix structure that reduces the rate of correct N-gram matching.

Benchmarking and assessing the performance of Arabic stemmers

Previous studies on the stemming of the Arabic language lack fair evaluation, full description of algorithms used or access to the source code of the stemmers and the datasets used to evaluate such stemmers. Freeing source codes and datasets is an essential step to enable researchers to enhance stemmers currently in use and to verify the results of these studies. This study laid the foundation of establishing a benchmark for Arabic stemmers and presents an evaluation of four heavy (root-based) stemmers for the Arabic language. The evaluation aims to assess the accuracy of each of the four stemmers and to show the strength of each. The four algorithms are: Al-Mustafa stemmer, Al-Sarhan stemmer, Rabab’ah stemmer and Taghva stemmer. The accuracy and strength tests used in this study ranked Rabab’ah stemmer as the first followed by Al-Sarhan, Al-Mustafa, and Taghva stemmers respectively.

The Effect of Stemming on Arabic Text Classification: An Empirical Study

The information world is rich of documents in different formats or applications, such as databases, digital libraries, and the Web. Text classification is used for aiding search functionality offered by search engines and information retrieval systems to deal with the large number of documents on the web. Many research papers, conducted within the field of text classification, were applied to English, Dutch, Chinese, and other languages, whereas fewer were applied to Arabic language. This paper addresses the issue of automatic classification or classification of Arabic text documents. It applies text classification to Arabic language text documents using stemming as part of the preprocessing steps. Results have showed that applying text classification without using stemming; the support vector machine SVM classifier has achieved the highest classification accuracy using the two test modes with 87.79% and 88.54%. On the other hand, stemming has negatively affected the accuracy, where the SVM accuracy using the two test modes dropped down to 84.49% and 86.35%.

An Approach for Arabic Text Categorization Using Association Rule Mining

Text Categorization (TC) has become one of the major techniques for organizing and managing online information. Several studies proposed the so-called associative classification for databases and few of these studies are proposed to classify text documents into predefined categories based on their contents. In this paper a new approach is proposed for Arabic text categorization. The approach facilitates the discovery of association rules for building a classification model for Arabic text categorization. An apriori based algorithm is employed for association rule mining. To validate the proposed approach, several experiments were applied on a collection of Arabic documents. Three classification methods using association rules were compared in terms of their classification accuracy; the methods are: ordered decision list, weighted rules, and majority voting. The results showed that the majority voting method is the best in most of experiments achieving an accuracy of up to 87%. On the other hand, the weigh...

A Multi-Label Classification Approach Based on Correlations Among Labels

Multi label classification is concerned with learning from a set of instances that are associated with a set of labels, that is, an instance could be associated with multiple labels at the same time. This task occurs frequently in application areas like text categorization, multimedia classification, bioinformatics, protein function classification and semantic scene classification. Current multi-label classification methods could be divided into two categories. The first is called problem transformation methods, which transform multi-label classification problem into single label classification problem, and then apply any single label classifier to solve the problem. The second category is called algorithm adaptation methods, which adapt an existing single label classification algorithm to handle multi-label data. In this paper, we propose a multi-label classification approach based on correlations among labels that use both problem transformation methods and algorithm adaptation methods. The approach begins with transforming multi-label dataset into a single label dataset using least frequent label criteria, and then applies the PART algorithm on the transformed dataset. The output of the approach is multi-labels rules. The approach also tries to get benefit from positive correlations among labels using predictive Apriori algorithm. The proposed approach has been evaluated using two multi-label datasets named (Emotions and Yeast) and three evaluation measures (Accuracy, Hamming Loss, and Harmonic Mean). The experiments showed that the proposed approach has a fair accuracy in comparison to other related methods.

Rough Set Theory for Arabic Sentiment Classification

Recently, the web has been a major place where people interact and express their views and sentiments. Researchers were attracted to conduct further analysis on this rich content known as Sentiment Analysis, Sentiment Classification or Opinion Mining. Rough Set Theory is a mathematical tool that can be used for classification and analysis of uncertain, incomplete or vague information. It can be used to significantly reduce the dimensionality of the data without much loss in information content, which is achieved using the concept of Reduct. This paper focuses on investigating the use of the Rough Set theory approach for Arabic Sentiment Classification. This paper compares some approaches that have been proposed to find Reducts to classify Arabic tweeting reviews. The Rosetta toolkit is used for testing where two main Reduct approaches were applied: Johnson Reducer and Genetic-based reducer. We compared the results of the approaches using cross validation evaluation method. The results showed that Genetic reducer achieved 57% of accuracy, which outperformed Johnson Reducer. The paper concludes that the Rough Set based approach is applicable for sentiment analysis of Arabic text but further investigation is required to evaluate other Reduct generation methods.

Evaluation of discernibility matrix based reduct computation techniques

Rough set theory provides some principles that are used for data classification and knowledge reduction. Reduct is one of the main concepts that can be used for feature set reduction and for data classification. Finding the reduct set is computationally expensive for data sets with large number of attributes. Several heuristic approached have been proposed to extract reduct sets where some of the approached used the Discernibility Matrix (DM) concept to perform the reduct computation. In this paper the Johnson reduction algorithm and the Object Reduct using Attribute Weighting technique algorithm (ORAW) for reduct computation are evaluated. The two approaches aim at reducing the number of features in the dataset. To evaluate the two approaches several UCI standard datasets were used in the experiments. The results of the experiments showed that the ORAW approach gives better results in term of classification accuracy where the average classification accuracy over eight data sets achieved by the ORAW approach was 85.6%; while Johnson approach achieved 78.8% of accuracy. For further evaluation, the two approaches were compared with some other well known classification techniques.

A hybrid approach to detecting alerts in Arabic e-mail messages

Detecting alert e-mails received daily by millions of subscribers from online news providers is a relatively new area of research which falls within the e-mail filtering field of research. Alert e-mails may address government, political issues, breaking news, and criminal attacks. This article proposes a hybrid approach based on both the Graham statistical filter and rule-based filters to detect and filter Arabic alert e-mails. The approach is basically language-independent. To test the performance of the proposed approach, several experiments have been conducted using a set of 1500 Arabic messages related to criminal activities collected manually from some news websites such as Al-Jazeera Net and BBC Arabic news. The results showed that the proposed approach has achieved a competitive performance in terms of accuracy, precision, and F-measure, where about 87% of the messages tested have been correctly detected and filtered by the proposed filter.

Application of Rough Set-Based Feature Selection for Arabic Sentiment Analysis

Sentiment analysis is considered as one of the recent applications of text categorization that categories the emotions expressed in text as negative, positive, and natural. Rough set theory is a mathematical tool used to analyze uncertainty, incomplete information, and data reduction. Indiscernibility, reduct, and core are essential concepts in rough set theory that can be employed for data classification and knowledge reduction. This paper proposes to use the rough set-based methods for sentiment analysis to classify tweets that are written in the Arabic language. The paper investigates the application of the reduct concept of rough set theory as a feature selection method for sentiment analysis. This paper investigates four reduct computation techniques to generate the set of reducts. For classification purposes, two rule generation algorithms have been studied to build the rough set rule-based classifier. An Arabic data set of 4800 tweets is used in the experiments to validate the use of reduct computation for Arabic sentiment analysis. The results of the experiments showed that using rough set reducts techniques lead to different results and some of them can perform better than non-rough set classifier. The best classification accuracy rate was for rough set classifier using the full attribute weighting reduct generation algorithm which achieved an accuracy of 74%. The primary results indicate that using the rough set theory framework for sentiment analysis is an appealing option where it can enhance the overall accuracy and reduce the number of used terms for classification which in turn will lead to a faster classification process, especially with a large dataset.

An associative rule-based classifier for Arabic medical text

Text classification is one of the methods used for managing, organising and retrieving the needed data among the huge available text. Several methods have been proposed to manipulate the text classification problem. In recent years, some studies proposed the use of Associative Classification AC approach. This paper examines an associative classification approach for the categorisation of text typed in Arabic language and related to medical domain. The approach discovers a set of association rules to build a classification model where three steps were applied to build the model: generating association rules, rule ordering and pruning, and then validation. The results of the experiments showed that the ordered decision list approach outperforms other approaches with accuracy reaching 90.6%. In general, the results of the experiments showed that association rule mining is a suitable method for building good classification models to categorise Arabic medical text.