طراحي پيكره زبان‌آموز دوزبانه ارمني ـ فارسي داراي برچسب خطاي املائي

The present study aims to design a bilingual Armenian–Persian learner corpus tagged with spelling error. Given the existing gap in bilingual corpus resources, this research seeks to identify patterns of spelling errors an‎d provide a data-driven infrastructure for linguistic an‎d computational analysis. The data for this research were collected from the written productions of Armenian learners in elementary an‎d secondary school levels at the "Armen" school, across various educational stages, an‎d were tagged using the INCEpTION software. To eva‎luate the validity of the corpus an‎d to perform intelligent data analysis, machine learning algorithms, including Ran‎dom Forest an‎d Support Vector Machine (SVM), were employed for the automatic classification of errors an‎d the assessment of corpus features. The final analysis of 1,116 spelling error tags showed that the category "main markers" had the highest frequency, accounting for 50% of the total data, which reflects the fundamental challenges of these learners at the phonological an‎d orthographic levels. Additionally, "multi-letter combinations" were identified as the second most frequent category, indicating the complexity of letter combinations for this group of learners. The results from the machine learning models also confirmed the high efficiency an‎d accuracy of these patterns in detecting an‎d automatically classifying errors. The corpus generated in this study, in addition to its application in refining bilingual school curricula, provides a valuable resource for developing natural language processing tools an‎d intelligent spell-checking systems.

5

163513