Religious texts have long influenced cultural, moral, and ethical systems, and have shaped societies for generations. Scriptures like the Bible, the Quran, and the Bhagavad Gita offer insights into fundamental human values and societal norms. Analyzing these texts with advanced methods can help improve our understanding of their significance and the similarities or differences between them. This study uses Natural Language Processing (NLP) techniques to examine these religious texts. Latent Dirichlet Allocation (LDA) is used for topic modeling to explore key themes, while GloVe embeddings and Sentence Transformers are used to compare topics between the texts. Sentiment analysis using Valence Aware Dictionary and sEntiment Reasoner (VADER) assesses the emotional tone of the verses, and corpus distance measurement is done to analyze semantic similarities and differences. The findings reveal unique and shared themes and sentiment patterns across the Bible, the Quran, and the Bhagavad Gita, offering new perspectives in computational religious studies.

This paper tries to decipher messages from the Quran and the Bible’s Mandarin translation using the multidimensional factor analysis (MDA) approach. Part-of-speech and word-meaning annotations were employed for data tagging. Seven syntactic and six semantic factors derived from the tagging systems demonstrated how the two scriptures are interpreted on the factor score scales. The analyses indicated that both holy books uphold a “persuade” and “preach” style with higher frequencies of imperative, advocative, and explanatory expressions. In addition, both favor the “interpersonal, non-numeric, and indicative” strategies to impress followers and practitioners alike with more elaborative wordings. The factor analysis approach also revealed that the Bible differs from the Quran by adopting more “motion, direction, and transportation” information, reflecting the deviation in their historical and religious backgrounds.

Zoroastrianism, one of the earliest known religions, reached its height of influence during the Sassanian period, embedding itself within the governmental structure before the rise of Islam in the 7th century led to a significant shift. Subsequently, a substantial body of Zoroastrian literature in Middle Persian (Pahlavi) emerged, primarily addressing religious, ethical, and legal topics and reflecting Zoroastrian responses to evolving Islamic jurisprudence. The text Šāyist nē šāyist (Licit and Illicit), which is central to this study, provides guidance on purity and pollution, offering insights into Zoroastrian legal principles during the late Sassanian period. This study marks the first known application of machine processing to Book Pahlavi texts, focusing on a jurisprudential Zoroastrian text. A Pahlavi corpus was compiled, and word embedding techniques were applied to uncover semantic relationships within the selected text. Given the lack of digital resources and data standards for Pahlavi, a unique dataset of vocabulary pairs was created for evaluating embedding models, allowing for the selection of optimal methods and hyperparameter settings. By constructing a complex network using these embeddings, and leveraging the scarcity of texts in this field, we used complex network analysis to extract additional information about the features of the text. We applied this approach to the chapters of the Šāyist nē šāyist book, uncovering more insights from each chapter. This approach facilitated the initial semantic analysis of Pahlavi legal concepts, contributing to the computational exploration of Middle Persian religious literature.

This study examines the use of Natural Language Processing (NLP) technology within the Islamic domain, focusing on developing an Islamic neural retrieval model. By leveraging the robust XLM-R base model, the research employs a language reduction technique to create a lightweight bilingual large language model (LLM). Our approach for domain adaptation addresses the unique challenges faced in the Islamic domain, where substantial in-domain corpora exist only in Arabic while limited in other languages, including English. The work utilizes a multi-stage training process for retrieval models, incorporating large retrieval datasets, such as MS MARCO, and smaller, in-domain datasets to improve retrieval performance. Additionally, we have curated an in-domain retrieval dataset in English by employing data augmentation techniques and involving a reliable Islamic source. This approach enhances the domain-specific dataset for retrieval, leading to further performance gains. The findings suggest that combining domain adaptation and a multi-stage training method for the bilingual Islamic neural retrieval model enables it to outperform monolingual models on downstream retrieval tasks.

Large Language Models (LLM) can be useful tools for translating Islamic literature written in Arabic into several languages, making this complex task technologically feasible, providing high-quality translations, at low cost and high-speed production enabled by parallel computing. We applied LLM-driven translation automation on a diverse corpus of Islamic scholarly works including: the Qur’an, Quranic exegesis (Tafseer), Hadith, and Jurisprudence from the Al-Shamela library. More than 250,000 pages have been translated into English, emphasizing the potential of LLMs to cross language barriers and increase global access to Islamic knowledge. OpenAI’s gpt-4o-mini model was used for the forward translation from Arabic to English with acceptable translation quality. Translation quality validation was achieved by reproducing Arabic text via back-translation from English using both the OpenAI LLM and an independent Anthropic LLM. Correlating the original source Arabic text and the back-translation Arabic text using a vector embedding cosine similarity metric demonstrated comparable translation quality between the two models.

The proliferation of Quranic content on digital platforms, including websites and social media, has brought about significant challenges in verifying the authenticity of Quranic verses. The inherent complexity of the Arabic language, with its rich morphology, syntax, and semantics, makes traditional text-processing techniques inadequate for robust authentication. This paper addresses this problem by leveraging state-of-the-art transformer-based Language models tailored for Arabic text processing. Our approach involves fine-tuning three transformer architectures BERT-Base-Arabic, AraBERT, and MarBERT on a curated dataset containing both authentic and non-authentic verses. Non-authentic examples were created using sentence-BERT, which applies cosine similarity to introduce subtle modifications. Comprehensive experiments were conducted to evaluate the performance of the models. Among the three candidate models, MarBERT, which is specifically designed for handling Arabic dialects demonstrated superior performance, achieving an F1-score of 93.80%. BERT-Base-Arabic also showed competitive F1 score of 92.90% reflecting its robust understanding of Arabic text. The findings underscore the potential of transformer-based models in addressing linguistic complexities inherent in Quranic text and pave the way for developing automated, reliable tools for Quranic verse authentication in the digital era.

This paper introduces the Morphological and Syntactical analysis for the Quran text. In this research we have constructed the MASAQ dataset, a comprehensive resource designed to address the scarcity of annotated Quranic Arabic corpora and facilitate the development of advanced Natural Language Processing (NLP) models. The Quran, being a cornerstone of classical Arabic, presents unique challenges for NLP due to its sacred nature and complex linguistic features. MASAQ provides a detailed syntactic and morphological annotation of the entire Quranic text that includes more than 131K morphological entries and 123K instances of syntactic functions, covering a wide range of grammatical roles and relationships. MASAQ’s unique features include a comprehensive tagset of 72 syntactic roles, detailed morphological analysis, and context-specific annotations. This dataset is particularly valuable for tasks such as dependency parsing, grammar checking, machine translation, and text summarization. The potential applications of MASAQ are vast, ranging from pedagogical uses in teaching Arabic grammar to developing sophisticated NLP tools. By providing a high-quality, syntactically annotated dataset, MASAQ aims to advance the field of Arabic NLP, enabling more accurate and more efficient language processing tools. The dataset is made available under the Creative Commons Attribution 3.0 License, ensuring compliance with ethical guidelines and respecting the integrity of the Quranic text.

This paper introduces the Hadith Simplification Dataset, a novel resource comprising 250 pairs of Classical Arabic (CA) Hadith texts and their simplified Modern Standard Arabic (MSA) equivalents. Addressing the lack of resources for simplifying culturally and religiously significant texts, this dataset bridges linguistic and accessibility gaps while preserving theological integrity. The simplifications were generated using a large language model and rigorously verified by an Islamic Studies expert to ensure precision and cultural sensitivity. By tackling the unique lexical, syntactic, and cultural challenges of CA-to-MSA transformation, this resource advances Arabic text simplification research. Beyond religious texts, the methodology developed is adaptable to other domains, such as poetry and historical literature. This work underscores the importance of ethical AI applications in preserving the integrity of religious texts while enhancing their accessibility to modern audiences.

This paper investigates the relationship between syntax and morphology in natural languages, focusing on the relation between the amount of information stored by word structure on the one hand, and word order on the other. In previous work, a trade-off between these was observed in a large corpus covering over a thousand languages, suggesting a dynamic ‘division of labor’ between syntax and morphology, as well as yielding proof for the efficient coding of information in language. In contrast, we find that the trade-off can be explained by differing conventions in orthographic word boundaries. We do so by redefining word boundaries within languages either by increasing or decreasing the domain of wordhood implied by orthographic words. Namely, we paste frequent word-pairs together and split words into their frequently occurring component parts. These interventions yield the same trade-off within languages across word domains as what is observed across languages in the orthographic word domain. This allows us to conclude that the original claims on syntax-morphology trade-offs were spurious and that, more importantly, there does not seem to exist a privileged wordhood domain where within- and across-word regularities yield an optimal or optimized amount of information.