This paper proposes a novel hybridization of two metaheuristic algorithms to solve the real-parameter single objective numerical optimization problems. The proposed Three-stage Cyclic Hybrid SFS and L-SHADE (3-sCHSL) algorithm integrates the high-level interactions of Stochastic Fractal Search (SFS) and L-SHADE algorithms in a single framework. Furthermore, a guided control population initialization strategy is injected in 3-sCHSL to address the stagnation and diversity loss issues as the evolution process evolves. The performance of the proposed algorithm is tested under different complexity levels of different transformations of the CEC 2021 benchmark suite with dimension 20. The experimental results demonstrated the efficiency and competitiveness of the proposed algorithm against recent state-of-the-art algorithms. In addition, 3-sCHSL achieved a superior performance when evaluating two engineering design problems.

Machine reading comprehension (MRC) requires machines to read and answer questions about a given text. This can be achieved through either predicting answers or extracting them. Extracting answers from text involves predicting the first and last index of the answer span within the paragraph. Training machines to answer questions requires datasets that are created for such a purpose. The lack of availability of benchmarking datasets for the Arabic language has hindered research into machine reading comprehension from Arabic text. The aim of this article is to propose an Arabic Span-Extraction-based Reading Comprehension Benchmark (ASER) and complement it with neural baseline models for performance evaluations. Detailed steps are depicted for building and evaluating ASER, which is an Arabic dataset created manually for the task of machine reading comprehension. It contains 10,000 records from different domains and is divided into training and testing sets. The results of ASER evaluation led to the conclusion that it is a challenging benchmark since the answers have varying lengths and human performance resulted in an exact match of 42%. On the other hand, two main baseline models were the focus of ASER experimentation: the sequence-to-sequence (Seq2Seq) model with different neural networks and the bidirectional attention flow (BIDAF) model. These experiments were implemented using different embeddings, and the results showed an exact match with lower values than human performance.

Real-life problems can be expressed as optimization problems. These problems pose a challenge for researchers to design efficient algorithms that are capable of finding optimal solutions with the least budget. Stochastic Fractal Search (SFS) proved its powerfulness as a metaheuristic algorithm through the large research body that used it to optimize different industrial and engineering tasks. Nevertheless, as with any meta-heuristic algorithm and according to the “No Free Lunch” theorem, SFS may suffer from immature convergence and local minima trap. Thus, to address these issues, a popular Differential Evolution variant called Success-History based Adaptive Differential Evolution (SHADE) is used to enhance SFS performance in a unique three-phase hybrid framework. Moreover, a local search is also incorporated into the proposed framework to refine the quality of the generated solution and accelerate the hybrid algorithm convergence speed. The proposed hybrid algorithm, namely eMpSDE, is tested against a diverse set of varying complexity optimization problems, consisting of well-known standard unconstrained unimodal and multimodal test functions and some constrained engineering design problems. Then, a comparative analysis of the performance of the proposed hybrid algorithm is carried out with the recent state of art algorithms to validate its competitivity.

Treebanks are valuable linguistic resources that include the syntactic structure of a language sentence in addition to part-of-speech tags and morphological features. They are mainly utilized in modeling statistical parsers. Although the statistical natural language parser has recently become more accurate for languages such as English, those for the Arabic language still have low accuracy. The purpose of this article is to construct a new Arabic dependency treebank based on the traditional Arabic grammatical theory and the characteristics of the Arabic language, to investigate their effects on the accuracy of statistical parsers. The proposed Arabic dependency treebank, called I3rab, contrasts with existing Arabic dependency treebanks in two main concepts. The first concept is the approach of determining the main word of the sentence, and the second concept is the representation of the joined and covert pronouns. To evaluate I3rab, we compared its performance against a subset of Prague Arabic Dependency Treebank that shares a comparable level of details. The conducted experiments show that the percentage improvement reached up to 10.24% in UAS and 18.42% in LAS.

Measuring semantic similarity between short texts is an important task in many applications of natural language processing, such as paraphrasing identification. This process requires a benchmark of sentence pairs that are labeled by Arab linguists and considered a standard that can be used by researchers when evaluating their results. This research describes an Arabic paraphrasing benchmark to be a good standard for evaluation algorithms that are developed to measure semantic similarity for Arabic sentences to detect paraphrasing in the same language. The transformed sentences are in accordance with a set of rules for Arabic paraphrasing. These sentences are constructed from the words in the Arabic word semantic similarity dataset and from different Arabic books, educational texts, and lexicons. The proposed benchmark consists of 1,010 sentence pairs wherein each pair is tagged with scores determining semantic similarity and paraphrasing. The quality of the data is assessed using statistical analysis for the distribution of the sentences over the Arabic transformation rules and exploration through hierarchical clustering (HCL). Our exploration using HCL shows that the sentences in the proposed benchmark are grouped into 27 clusters representing different subjects. The inter-annotator agreement measures show a moderate agreement for the annotations of the graduate students and a poor reliability for the annotations of the undergraduate students.

In this paper, an ensemble of machine learning classifiers approach is used to classify the sentiment polarity of Arabic text. This approach is based on the majority voting algorithm in conjunction with four classifiers, namely Naive Bayes, Support Vector Machines, Decision Trees and K-Nearest Neighbor algorithms. Four combinations of these classifiers are formed and three classifiers are chosen for each voting combination. The performance of each classifier is evaluated and compared to ensemble voting combination performance. Different experiments have been performed to evaluate unigram and bigram features. Three datasets with different sizes are used in our experiments. The first dataset contains 500 movie reviews, the second one contains 2000 Arabic tweets and the third one contains 16448 of Arabic book reviews. The experimental results show that the ensemble of the classifiers comparatively gives better results than individual classifiers. They also reveal that the support vector machine classifier outperforms the other individual classifiers. Moreover, the results of the bigram feature are better than the results of the unigram feature.

The rapid growth of the Internet and other computing facilities in recent years has resulted in the creation of a large amount of text in electronic form, which has increased the interest in and importance of different automatic text processing applications, including keyword extraction and term indexing. Although keywords are very useful for many applications, most documents available online are not provided with keywords. We describe a method for extracting keywords from Arabic documents. This method identifies the keywords by combining linguistics and statistical analysis of the text without using prior knowledge from its domain or information from any related corpus. The text is preprocessed to extract the main linguistic information, such as the roots and morphological patterns of derivative words. A cleaning phase is then applied to eliminate the meaningless words from the text. The most frequent terms are clustered into equivalence classes in which the derivative words generated from the same root and the non-derivative words generated from the same stem are placed together, and their count is accumulated. A vector space model is then used to capture the most frequent N-gram in the text. Experiments carried out using a real-world dataset show that the proposed method achieves good results with an average precision of 31% and average recall of 53% when tested against manually assigned keywords.