AI translation features of medieval persian texts

The target audience of this article is the few specialists currently engaged in the professional translation of medieval scholarly and religious texts from Oriental languages (Persian, Arabic, Turkic). Using the Persian language as an example, this paper briefly reviews all modern types of «non-human» translation of these texts: computer-assisted translation (CAT); neural machine translation (NMT) involving a set of microchips and algorithms, which today is referred to as artificial intelligence (AI); generative AI translation (GenAI); and AI translation based on large language models (LLMs). When using any of these methods, to a greater or lesser extent, a translator needs to create a glossary of key terms and words with their unambiguous target language equivalents, i.e., a terminological database of a specific author for a specific work. The preparation of such a database is practical in two cases: 1) in the presence of a voluminous text (or texts) of a single thematic focus (preferably by the same author), wherein a glossary prepared during initial phase substantially optimizes translation efficiency in subsequent stages; 2) during interruptions in work due to objective and subjective reasons, where a previously prepared (or generated) and subsequently forgotten glossary facilitates the resumption of text translation within the same thematic framework.