This study provides a reproducible, corpus-based framework for the structural and semantic analysis of educational terminology related to information technology in a bilingual (English-Uzbek) context. A comprehensive bilingual corpus containing approximately 1.2 million tokens (620,000 in English and 580,000 in Uzbek) from information technology textbooks and educational resources was constructed. A carefully selected and manually annotated evaluation set of 5,000 lexical items from this corpus served as the basis for quantitative evaluation of the extraction and classification models. Terminology candidates were automatically extracted from the full corpus using a hybrid statistical and linguistic algorithm combining TF-IDF, C-value/NC-value, and morphosyntactic filtering. Evaluation on a carefully selected set showed high reliability: precision = 0.91, recall = 0.87, and F1 score = 0.89. A prototype bilingual matching tool using FastText subword embeddings / vectors and cosine similarity achieved an accuracy of 0.86 for matching English and Uzbek terms. In addition, a supervised XGBoost classifier was trained on the evaluation set to detect neologisms and metaphorical extensions of computer terminology, achieving an F1 score of 0.88. The proposed system provides a repeatable natural language processing pipeline for extracting, matching, and semantic classification of bilingual terminology. It serves as both a quantitative benchmark and a practical tool for integrating computer vocabulary into educational resources. This approach facilitates the systematic study of the structural and semantic dynamics of bilingual educational lexicons, thereby supporting future research and the development of practical curricula.