Abstract
Text classification is the process of assigning a text or a document to various predefined classes or categories to reflect their contents. With the rapid growth of Arabic text on the Web, studies that address the problems of classification and segmentation of the Arabic language are limited compared to other languages, most of which implement word-based and feature extraction algorithms. This paper adopts a PPM character-based compression scheme to classify and segment Classical Arabic (CA) and Modern Standard Arabic (MSA) texts. An initial experiment using the PPM classification method on samples of text resulted in an accuracy of 95.5%, an average precision of 0.958, an average recall of 0.955 and an average F-measure of 0.954, using the concept of minimum cross-entropy. PPM-based classification experiments on standard Arabic corpora showed that they contained different types of text (CA or MSA), or a mixture of the both (CA and MSA). Further experiments with the same corpora showed that a more accurate picture of the contents of the corpora was possible using the PPM-based segmentation method. Tag-based compression experiments (using tags produced by parts-of-speech Arabic taggers) also showed that the quality of the tagging (as measured by compression quality) is significantly affected when tagging either CA and MSA text. The conclusion is that NLP applications (such as taggers) should treat these texts separately and use different training data for each or process them differently.
Highlights
Text classification is the process of automatically assigning a document to different predefined classes or categories to reflect their contents [1]
Classification of Classical Arabic (CA) and Modern Standard Arabic (MSA) text was performed on sample texts using a Prediction-by-Partial Matching (PPM) character-based compression scheme achieving an accuracy of 95.5%, an average precision of 0.958, an average recall of 0.955 and an average F-measure of 0.954
A classification of Arabic corpora was performed and the results showed that different sub-genres of some Arabic corpora contain different types of Arabic text since the compression size for other corpora indicated that the texts were a mixture between CA and MSA
Summary
Text classification is the process of automatically assigning a document to different predefined classes or categories to reflect their contents [1]. The work in this paper instead uses an approach based on the Prediction-by-Partial Matching (PPM) compression scheme as the basis of both text classification and segmentation This Markov-based approach effectively uses character-based language models and has been employed in many NLP tasks in the past often with state-of-the-art results or results competitive with traditional schemes [1], [31]–[36]. The use of Markovbased approximations standard in character-based language modelling avoids the issue of explicit feature selection that is applied in traditional classification and segmentation algorithms which may discriminate some important features of the text [1], [37]. The segmentation process performed in this study applies a Viterbi-style algorithm which produces an accurate estimate of each class, category or topic located in the text [34].
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
More From: International Journal of Advanced Computer Science and Applications
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.