Dr. Anya Sharma, a computational linguist, is developing an automated script identification system for a digital archive containing historical documents from various regions of the Ottoman Empire. The archive includes texts in Ottoman Turkish, Arabic, Persian, and various Balkan languages, all potentially using different script variants or even code-switching between scripts within the same document. Her initial system relies solely on algorithmic script detection based on character frequency analysis. However, she notices that the system frequently misidentifies the dominant script in documents containing Ottoman Turkish, particularly when the text includes a high proportion of Arabic loanwords or regional script variations common in the Balkans. Considering the complexities of script usage within the Ottoman Empire and the limitations of purely algorithmic approaches, which of the following strategies would MOST effectively improve the accuracy of Dr. Sharma\'s script identification system?