A Study on Knowledge Discovery from Multilingual Educational Corpora Using Artificial Intelligence Techniques

Abstract

Multilingual educational materials include rich information which is hard to analyze due to their multilinguality, computational limitation, and the complexity related to low-resource languages. This research will introduce a knowledge discovery system that exploits artificial intelligence technologies to extract relevant patterns out of multilingual educational texts through natural language processing, educational data mining, topic modeling, and knowledge graphs. This proposed system uses lexical, semantic, and context representation based on term frequency-inverse document frequency, word embeddings, and multilingual transformers respectively to facilitate the analysis of cross-lingual educational materials. This study designs a multilingual processing pipeline including language detection, normalization, tokenization, morphological processing, concept extraction, and knowledge fusion for educational text mining in Uzbek, Russian, and English texts. This system allows automatic discovery of topics, classification of grade levels, aligning of curriculums, and representation of educational knowledge in a structured manner. The hybrid fusion model achieved an accuracy of 97.3% and an F1-score of 96.9%, while the cross-lingual fusion component improved concept matching accuracy to 92.7%. It is proven by the experiments conducted on this model that integration of representation methods improves classification, semantic, and concept consistency. The knowledge graph module allows capturing the relationship between educational concepts, and cross-language fusion helps to minimize inconsistencies in multilingual learning content. The proposed system offers an effective solution for intelligent educational analytics and multilingual curriculums management especially in low-resource languages.