نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی دکتری، گروه علم اطلاعات و دانش‌شناسی، دانشکده روانشناسی و علوم تربیتی، دانشگاه خوارزمی، تهران، ایران.
2 استادیار، گروه علم اطلاعات و دانش‌شناسی، دانشکده روانشناسی و علوم تربیتی، دانشگاه خوارزمی، تهران، ایران
3 استادیار، گروه مهندسی کامپیوتر، دانشکده فنی و مهندسی، دانشگاه قم، قم، ایران.
4 استادیار، گروه مهندسی کامپیوتر، دانشکده مهندسی برق و کامپیوتر، دانشگاه خوارزمی، تهران، ایران.
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Purpose: The present research aims to develop a model for automatically inferring the relationships between terms in the Thesaurus of Islamic Sciences using Graph Convolutional Networks (GCN). By employing new algorithms in the field of deep learning, the research seeks to enhance the efficiency of information retrieval in the Thesaurus of Islamic Sciences. To enhance accuracy and comprehensiveness, reduce costs, and improve relationships between terms.
Method: The current research employed used of convolutional networks method, networks, is are one of the crucial techniques methods in the field of learning. This method is capable of leveraging from the relationship patterns in the while also focusing on to the characteristics of each node. The dataset under study comprises all the terms from the thesaurus of Islamic sciences generated between 1994 and the early 2022, which are represented as a graph. The vertices represent the terms, and the edges represent the relationships between the terms in the graph. This graph is provided as input to the convolutional network, which then generates a model for the automatic inference of connections. And in order to analyze the obtained outputs, AP and ROC standards have been used.
Findings: The revealed showed the model achieved the average accuracy 75% and a Roc score of 72% obtained for the data. It is noteworthy to accept the results considering that this method was used for the first time in the field of Islamic sciences and thesauruses.
Conclusion: Despite shift in preference opinion thesauri thesauruses to ontologies, the use thesauri remains still of particularly especially in Iran. Compared to previous research, the method used to construct the thesaurus is different, resulting in more reliable outcomes. Consequently, we can expect improved results for various purposes, such as automatic indexing. New advancements in natural language processing and deep learning also give us hope for improvements in information retrieval and automatic indexing.
کلیدواژه‌ها [English]
)
ارسال نظر در مورد این مقاله