Methods of the OntologyAutomatic Learning in the Fields of Quranic Concepts: a Scoping Review Study

Document Type : Original Article


1 Assistant Professor, Department of Knowledge Dissemination, Islamic Sciences and Culture Academy, Qom,

2 Assistant Professor, Department of Knowledge and Information Science, University of Isfahan, Isfahan, Iran

3 PhD. Student, Knowledge and Information Science, Shahid Chamran University, Ahvaz, Iran

4 PhD. Student, Medical Library and Information Science, Iran University of Medical Sciences, Tehran, Iran



Objective: Today, semantic technology offers a new approach in organizing Quranic knowledge with the aim of providing meaningful information and representing Quranic teachings. Ontologies are a tool to formally express concepts and relationships in a specific domain. In the same way, the development of ontology as a tool for representing the effulgence and extracting the knowledge of the Quran is not only valuable, but also necessary. Ontology learning and its methods automatically to extract concepts are important topics in the field of Semantic Web and its technologies. Recently, the development and application of ontologies learning for the extraction of Quranic concepts has been considered. Therefore, the aim of the current research is to comprehensively investigate the ontologies automatic learning in the field of extracting knowledge and Quranic concepts in order to clarify the current and future situation. The investigated criteria were data set, learning methods, evaluation methods, results and future suggestions of studies in the field of ontologies automatic learning of the Quran.
Methodology: The research was conducted by the scoping review method in accordance with PRISMA guidelines and based on Arksey & O’Malley procedure. This process describes a protocol for matching the results of existing studies with research questions and criteria. The five steps suggested by Arksey & O’Malley are as follows: 1. Identify and design the research question(s), 2. Conduct search strategies advocate for relevant studies through the selection of appropriate keywords and Boolean operators, 3. Final selection of relevant studies, considering the inclusion and exclusion criteria, 4. Tabulating the data, and finally, 5. Reporting its results. Sources were searched in seven scientific databases including Emerald, Science Direct, IEEE Xplore Digital Library, Google Scholar, Web of Science, and Scopus. The search process has been done in April 2023. A number of 811 articles, regardless of the time limit, were evaluated and selected. In order to organize the retrieved articles, EndNote resource management software was used and after matching the titles in different databases, 317 duplicate articles were removed. After reviewing the abstracts, the entry and exit criteria and the quality of the articles were applied. Also, in order to avoid bias in the selection of articles, during a random review, two independent researchers in the field of ontology automatic learning were evaluated and finally 25 articles were selected as review criteria.
Findings: Most of the study in the field of Quranic data set were in English and Arabic languages, and most of them used the English translation of Al-Hilali and Khan's Quran. The use of a limited data set was the most important limitation of the research conducted in the field of automatic learning of Quranic ontologies. Most of the studies have used normalization methods, text clustering and categorization, text summarization, information extraction, similarity and finding famous entities. Of course, in some studies, artificial intelligence methods such as neural network have also been used. In addition, the findings showed that data mining algorithms based on statistics and probability methods for learning and constructing automatic ontologies was apparently surging in popularity among researchers. Evaluation methods includes calculating accuracy, recall and F criteria in the application of automatic learning algorithms in Quranic ontologies. The studies that have used artificial intelligence techniques, by Semantic analysis, inference, modeling and validation of inferred data have achieved results such as sound recognition for teaching Quran reading, recognition of literary arrays and creating thematic connections in Quranic concepts as well as creating connections between these concepts and concepts in other religions. The evaluation of the presented methods for ontology automatic learning shows that the combined use of data mining methods and artificial intelligence brings better results. Most of the results of this field are in two general categories. The first category was based on the use of data mining, text mining and machine learning methods to automatically extract three concepts and dimensions (subject-predicate-object) along with Semantic relationships from the text of the Quran. The other category compares the performance of methods and algorithms based on statistics and similarity, such as TF, TF-IDF, AVE-TF, Ridf, TIM, N-gram, FREyA, Pos Taggin, Levenshtein, Log Likelihod, Herset, etc. in extracting concepts for the construction of the Quranic ontologies. The findings of the future studies review show the researchers' interest in artificial intelligence algorithms and their use in ontology learning and the automatic and semi-automatic development of Quranic ontologies. The lack of correct data sets is the reason for the inability of the world's advanced artificial intelligence systems such as GPT 4, which must be addressed in the future.
Discussion and conclusion: The results of this study can help to direct future research about the best practices in the automatic development of Quranic ontologies. This issue can be taken into consideration by designing a comprehensive Quranic ontology that covers all topics and concepts according to the context of the Quran, and by creating a comprehensive ontology of the Quranic concepts, it will guide users towards the retrieval of Quranic knowledge. Also, more use of artificial intelligence and natural language processing methods, such as GPT as a machine learning model for natural language text generation by deep neural network, it seems essential in the development of automatic learning of Quranic ontologies. Machine learning requires the existence of big data in the field of the Qur'an, hence the creation of standard data sets is one of the future studies.


Main Subjects