Speakers
Description
This paper presents a pilot study exploring the use of artificial intelligence in reversing a bilingual dictionary. The Modern Icelandic-English Dictionary is a freely available online resource. The aim was to transform this Icelandic-English dictionary into a functional English-Icelandic dictionary while minimising manual labour. Using the DANTE English lexical database as a reference, the study focused on English words beginning with the letter m. The process was divided into five phases. First, SQL queries were used to generate an initial reverse listing. Second, the Gemini large language model was employed to process the output and combine it with data from the DANTE list. The resulting dataset contained 4,750 potential entries, of which 1,750 lacked an Icelandic equivalent. In the third phase, Gemini supplied missing translations, while in the fourth phase it performed sense disambiguation for polysemous entries. The fifth phase involved post-processing, including the elimination of unnecessary entries and extensive proofreading. The findings suggest that AI can successfully handle three of the five phases. However, substantial post-processing still requires human expertise, although AI might also be used to facilitate parts of that work.