Arabic is among the most challenging languages in the world. Unfortunately, the scarcity of Arabic datasets makes studies in Arabic speech technology demanding. This paper introduces SADA, the Saudi Audio Dataset for Arabic, with 668 hours of high-quality audio suitable for supervised training. The audio recordings were sourced from 57 television shows provided by the Saudi Broadcasting Authority. The audio covers both read and spontaneous speaking styles in various genres. The National Center for Artificial Intelligence in Saudi Arabia transcribed and prepared the data for training and processing. The recordings are in Arabic. Most are in Saudi dialects, while other Arabic dialects include Yemeni, Egyptian, and Levantine. The dataset is split into training, validation, and testing sets to enhance its usage. The validation and testing sets contain 10 hours of audio segments each. Besides giving a detailed description of the dataset, wide range of speech recognition experiments using standard tools are also presented.

@INPROCEEDINGS{10446243,
  author={Alharbi, Sadeen and Alowisheq, Areeb and Tüske, Zoltán and Darwish, Kareem and Alrajeh, Abdullah and Alrowithi, Abdulmajeed and Tamran, Aljawharah Bin and Ibrahim, Asma and Aloraini, Raghad and Alnajim, Raneem and Alkahtani, Ranya and Almuasaad, Renad and Alrasheed, Sara and Alsubaie, Shaykhah and Alonaizan, Yaser},
  booktitle={ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, 
  title={SADA: Saudi Audio Dataset for Arabic}, 
  year={2024},
  volume={},
  number={},
  pages={10286-10290},
  keywords={Training;Vocabulary;TV;Speech recognition;Speech processing;Standards;Testing;Arabic dataset;dialectal Arabic data},
  doi={10.1109/ICASSP48485.2024.10446243}
}

In this paper, we present a recipe for building a good Arabic-English neural machine translation. We compare neural systems with traditional phrase-based systems using various parallel corpora including UN, ISI and Ummah. We also investigate the importance of special preprocessing of the Arabic script. The presented results are based on test sets from NIST MT 2005 and 2012. The best neural system produces a gain of +13 BLEU points compared to an equivalent simple phrase-based system in NIST MT12 test set. Unexpectedly, we find that tuning a model trained on the whole data using a small high quality corpus like Ummah gives a substantial improvement (+3 BLEU points). We also find that training a neural system with a small Arabic-English corpus is competitive to a traditional phrase-based system.

@Article{alrajeh:18:1,
  author    = {Abdullah Alrajeh},
  title     = {A Recipe for {A}rabic-{E}nglish Neural Machine Translation},
  journal   = {Computing Research Repository},
  volume    = {arXiv:1808.06116},
  year      = {2018},
  url       = {https://arxiv.org/abs/1808.06116}
}

In state-of-the-art phrase-based statistical machine translation systems (SMT), modelling phrase reorderings is an important need to enhance naturalness of the translated outputs, particularly when the grammatical structures of the language pairs differ significantly. The challenge in developing machine learning methods for machine translation can be summarised in two points. First is the ability to characterise language features such as morphology, syntax and semantics. Second is adapting complex learning algorithms to process large corpora.

Posing phrase movements as a classification problem, we exploit recent developments in solving large-scale SVM, Multiclass SVM and Multinomial Logistic Regression. Using dual coordinate descent methods for learning, we provide a mechanism to shrink the amount of training data required for each iteration. Hence, we produce significant saving in time and memory while preserving the accuracy of the models. These efficient classifiers allow us to build large-scale discriminative reordering models. We also explore a generative learning approach namely naive Bayes. Our Bayesian model is shown to be superior to the widely-used lexicalised reordering model. It is fast to train and the storage requirement is many times smaller than the lexicalised model. Although discriminative models might achieve higher accuracy than naive Bayes, the absence of iterative learning is a critical advantage for very large corpora.

Our reordering models are fully integrated with the Moses machine translation system, widely used in the community. Evaluated in large-scale translation tasks, our models have proved successful for two very different language pairs: Arabic-English and German-English.

@phdthesis{alrajeh:thesis,
  author  = "Abdullah Alrajeh",
  title   = "Large-scale reordering models for statistical machine translation",
  school  = "The University of Southampton",
  year    = "2015",
}

In state-of-the-art phrase-based statistical machine translation systems, modelling phrase reorderings is an important need to enhance naturalness of the translated outputs, particularly when the grammatical structures of the language pairs differ significantly. Posing phrase movements as a classification problem, we exploit recent developments in solving large-scale multiclass support vector machines. Using dual coordinate descent methods for learning, we provide a mechanism to shrink the amount of training data required for each iteration. Hence, we produce significant computational saving while preserving the accuracy of the models. Our approach is a couple of times faster than maximum entropy approach and more memory-efficient (50% reduction). Experiments were carried out on an Arabic-English corpus with more than a quarter of a billion words. We achieve BLEU score improvements on top of a strong baseline system with sparse reordering features.

@article{pbml-103-alrajeh-niranjan,
  author  = "Abdullah Alrajeh and Mahesan Niranjan",
  title   = "Scalable Reordering Models for {SMT} based on Multiclass {SVM}",
  journal = "The Prague Bulletin of Mathematical Linguistics",
  volume  = "103",
  pages   = "65--84",
  url     = "http://ufal.mff.cuni.cz/pbml/103/art-alrajeh-niranjan.pdf",
  year    = "2015"
}

Stochastic gradient descent has been advanced as a computationally efficient method for large-scale problems. In classification problems, many proposed linear support vector machines are very effective. However, they assume that the data is already in memory which might be not always the case. Recent work suggests a classical method that divides such a problem into smaller blocks then solves the sub-problems iteratively. We show that a simple modification of shrinking the dataset early will produce significant saving in computation and memory. We further find that on problems larger than previously considered, our approach is able to reach solutions on top-end desktop machines while competing methods cannot.

@InProceedings{alrajeh:2014:ICMV,
  author    = {Alrajeh, Abdullah and Takeda, Akiko and Niranjan, Mahesan},
  title     = {Memory-efficient Large-scale Linear Support Vector Machine},
  booktitle = {Proceedings of SPIE: Seventh International Conference on Machine Vision (ICMV 2014)},
  volume    = {9445},
  pages     = {944527-944527-6},
  publisher = {SPIE},
  year      = {2015},
  month     = {February},
  address   = {Milan, Italy},
  doi       = {10.1117/12.2180925},
  url 	    = { http://dx.doi.org/10.1117/12.2180925},
}

Phrase reordering is a challenge for statistical machine translation systems. Posing phrase movements as a prediction problem using contextual features modeled by maximum entropy- based classifier is superior to the commonly used lexicalized reordering model. However, Training this discriminative model using large-scale parallel corpus might be computationally expensive. In this paper, we explore recent advancements in solving large- scale classification problems. Using the dual problem to multinomial logistic regression, we managed to shrink the training data while iterating and produce significant saving in computation and memory while preserving the accuracy.

@InProceedings{D14-1183,
  author    = "Alrajeh, Abdullah and Niranjan, Mahesan",
  title     = "Large-scale Reordering Model for Statistical Machine Translation using Dual Multinomial Logistic Regression",
  booktitle = "Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
  month     = "October",
  year      = "2014",
  publisher = "Association for Computational Linguistics",
  pages     = "1758--1763",
  address   = "Doha, Qatar",
  url       = "http://aclweb.org/anthology/D14-1183"
}

In phrase-based statistical machine translation systems, variation in grammatical structures between source and target languages can cause large movements of phrases. Modeling such movements is crucial in achieving translations of long sentences that appear natural in the target language. We explore generative learning approach to phrase reordering in Arabic to English. Formulating the reordering problem as a classification problem and using naive Bayes with feature selection, we achieve an improvement in the BLEU score over a lexicalized reordering model. The proposed model is compact, fast and scalable to a large corpus.

@InProceedings{alrajeh-niranjan:2014:W14-33,
  author    = {Alrajeh, Abdullah  and  Niranjan, Mahesan},
  title     = {Bayesian Reordering Model with Feature Selection},
  booktitle = {Proceedings of the Ninth Workshop on Statistical Machine Translation},
  month     = {June},
  year      = {2014},
  address   = {Baltimore, Maryland, USA},
  publisher = {Association for Computational Linguistics},
  pages     = {477--485},
  url       = {http://www.aclweb.org/anthology/W/W14/W14-3361}
}

Automated document classification is an important text mining task especially with the rapid growth of the number of online documents present in Arabic language. Text classification aims to automatically assign the text to a predefined category based on linguistic features. Such a process has different useful applications including, but not restricted to, e-mail spam detection, web page content filtering, and automatic message routing. This paper presents the results of experiments on document classification achieved on seven different Arabic corpora using statistical methodology. The performance of two popular classification algorithms in classifying the aforementioned corpora has been evaluated.

Electronically formatted Arabic free-texts can be found in abundance these days on the World Wide Web, often linked to commercial enterprises and/or government organizations. Vast tracts of knowledge and relations lie hidden within these texts, knowledge that can be exploited once the correct intelligent tools have been identified and applied. For example, text mining may help with text classification and categorization. Text classification aims to automatically assign text to a predefined category based on identifiable linguistic features. Such a process has different useful applications including, but not restricted to, E-Mail spam detection, web pages content filtering, and automatic message routing. In this paper an overview of King Abdulaziz City for Science and Technology (KACST) Arabic Text Classification Project will be illustrated along with some preliminary results. This project will contribute to the better understanding and elaboration of Arabic text classification techniques.

Technology has shown great promise in providing access to textual information for visually impaired people. Optical Braille Recognition (OBR) allows people with visual impairments to read volumes of typewritten documents with the help of flatbed scanners and OBR software. This project looks at developing a system to recognize an image of embossed Arabic Braille and then convert it to text. It particularly aims to build fully functional Optical Arabic Braille Recognition system. It has two main tasks, first is to recognize printed Braille cells, and second is to convert them to regular text. Converting Braille to text is not simply a one to one mapping, because one cell may represent one symbol (alphabet letter, digit, or special character), two or more symbols, or part of a symbol. Moreover, multiple cells may represent a single symbol.