Arabic text categorization is an important task in text mining particularly with the fast-increasing quantity of the Arabic online data. Deep neural network models have shown promising performance and indicated great data modeling capacities in managing large and substantial datasets. This article investigates convolution neural networks (CNNs), long short-term memory (LSTM) and their combination for Arabic text categorization. This work additionally handles the morphological variety of Arabic words by exploring the word embeddings model using position weights and subword information. To guarantee the nearest vector representations for connected words, this article adopts a strategy for refining Arabic vector space representations using semantic information embedded in lexical resources. Several experiments utilizing different architectures have been conducted on the OSAC dataset. The obtained results show the effectiveness of CNN-LSTM without and with retrofitting for Arabic text categorization in comparison with major competing methods.
A fast region-based shape descriptor, the disk harmonic moments descriptor (DHMD), is proposed. The authors propose a mapping of the 3D spherical harmonics transform (SHT) to the unit disc, and formulate a disc harmonic transform (DHT) based on a new set of orthogonal harmonic moments. Before applying the transform DHT, the shape undergoes a preprocessing in order to ensure shift and scaling invariance. The DHMD is compared to three sets of descriptors: one is formed by three SHT-based descriptors and the other two gather some well-known contour-based and region-based descriptors. The authors consider the set B of the MPEG-7-CE-1 database for experimental validation. To benchmark the performance of the compared descriptors, precision-recall pair is employed. A computational comparison is also proposed. The DHMD achieves faster and better retrievals and seems to be competitive with the cited descriptors. This improvement should encourage the use of DHMD features in different image retrieval systems that support huge databases.
This paper presents a novel 2D region-based shape descriptor inspired by 3D spherical harmonics. We propose an adaptation of the 3D spherical harmonics, which provides a novel set of orthogonal harmonic moments defined on a 2D unit disc. A novel generation of region-based shape descriptors is proposed: the disk harmonic moment descriptor (DHMD). This new descriptor is compared with other methods from the same category. Set B of the MPEG-7-CE-1-shape database is used for experimental validation. To benchmark the performance of the compared descriptors precision-recall pair and several statistics provided by the Princeton Shape Benchmark are employed. With its good accuracy in 2D shape retrieval, the proposed descriptor is very useful as a generic shape descriptor for content-based image retrieval applications.