Definiranje značenja bosanskih frazema kroz prizmu vještačke inteligencije

Predmet istraživanja u radu je definiranje značenja bosanskih frazema koje vrši ChatGPT, chatbot vještačke inteligencije. U tu svrhu prikupljen je korpus od 50 frazema – poredbenih i neporedbenih – čija je obavezna sastavnica leksema orijentalnog porijekla (iz arapskog, perzijskog ili turskog jezika). Analizom se želi utvrditi koliko prisustvo orijentalizma u frazemi doprinosi ili odmaže u definiranju značenja frazeme, s naročitim akcentom na orijentalizme specifične za bosanski jezik i one koji su općeprihvaćeni i zastupljeni i u hrvatskom i srpskom jeziku. Pored značenja frazema, analiziran je i kontekst njihove upotrebe radi utvrđivanja uspješnosti vještačke inteligencije da ponudi adekvatne primjere.

Research conducted on a corpus of 50 comparative and non-comparative phrasemes containing lexemes of oriental origin has shown that the quality of definitions provided by artificial intelligence (AI) is high for phrasemes with generally accepted and widely used orientalisms (70%) and for comparative phrasemes with orientalisms of transparent origin (80%). However, this result is overall poorer for phrasemes with orientalisms of transparent origin that are specific to the Bosnian language (56.66%). Nevertheless, when these findings are supplemented with results on partial semantic matches of Bosnian language phrasemes, a high level of accuracy in their definitions can be established. In addition, an adequate definition of phraseme meanings is accompanied by a suitable context of their usage. However, other parameters, such as glosses, synonyms, etc., lag significantly behind the meaning of phrasemes and the context of their usage, requiring increased caution from lexicographers and dictionary users when relying on this information.